Before examining children's use of referring expressions in the data gathered for this study, it is necessary to look in detail at three areas: at the means by which referents are established in conversation; at what is known about the developmental aspects of the establishment of referent; and at what has been observed of the difficulties experienced in this area by children with pragmatic problems.


In the Pragmatic Protocol Prutting defines the communicative parameter Specificity/Accuracy as follows:

"Lexical items of best fit considering the text. examples: Appropriate behaviours: the ability to be specific and make appropriate lexical choices to clearly convey information in the discourse. Inappropriate behaviours: overuse of unspecified referents that results in ambiguity of the message. Also includes inappropriate choice of lexical items that do not facilitate understanding." (Prutting and Kirchner, 1987, p. 118).

Specificity/Accuracy is thus concerned with referring, that is with "the process of using a linguistic expression to pick out for one's addressee an individual entity - a particular person, thing, place, event etc. - or particular set of entities." (Griffiths, 1979, p. 106). An entity which is picked out in this way is called a referent. Referents may be entities of any kind, including "people, objects, actions and processes, conditions, ideas and attributes." (Cole, 1982. p. 51). They may also be part of the discourse itself, as in, 'That's a rude word'. The referent is said to be unspecified or unestablished in instances where the referring expression does not establish the identity of the referent for the addressee.

Although the referent itself is an individual entity, as distinct from what has been called the 'compositional meaning' contained in a proposition, the referring expression may extend well beyond a single lexical item. For example, to specify a particular hat a speaker may use a variety of referring expressions, such as: "it", "your hat", "a very gorgeous hat', "the hat that you're going to wear if you go on the swing".

The expression used to establish referent may also extend beyond the verbal to the vocal or gestural. For example, Tanz (1980) has drawn attention to the way in which young children use gesture to disambiguate deictic terms. Deictic terms are "expressions which serve to direct the hearer's attention to spatial or temporal aspects of the situation of utterance". (Wales, 1986, p. 401). As noted above (1.5.2), gesture may sometimes be used not simply to disambiguate a verbal expression but as a complete substitute for one. For example, instead of a referring expression such as 'that' or 'there' being accompanied by a pointing gesture, the gesture may be used on its own.

Referring expressions fall into two broad categories: those where the referent introduced by the speaker is new to the addressee(s) and those where the identity of the referent can be recognised by the addressee(s) because it is already established within the context of the conversation. Referents of the latter type are described as 'given' and will be discussed first.



The identity of 'given' referents is specified by means of cohesive devices, the nature of which has been investigated in detail by Halliday and Hasan (1976; 1985) and by Hasan (1984). Cohesive devices are used to refer back to entities whose identity is already established (or 'given') between conversational partners. They thus fulfill a dual role. In their function as cohesive ties they bring about a continuity of reference which helps to make the discourse into a connected whole. In their function as referring expressions they specify particular referents for the conversational partner(s). Referring expressions of this type can only be interpreted by reference to some aspect of the context of the discourse, as illustrated by the initial terms in the three following examples.

1. She works hard. (pronominal.)

2. There are my keys. (demonstrative.)

3. The pets were frightened. (definite article.)

One means by which such expressions may be interpreted is by reference to the linguistic context (endophoric reference). The connection is most commonly anaphoric, that is to say an expression refers back to a referent which is already mutually specific, as in: 'My mother looks tired. She works hard.' It may also, however, be cataphoric, referring the addressee forward to a referent which is still to be established, as in, 'She works hard, my mother.' As already indicated, a referent may also be a feature of the discourse itself ( e.g. 'That doesn't rhyme').

A referring expression which relies for its interpretation on observation of the physical context of the utterance is said to be exophoric. If, for example, in saying 'She works hard' the speaker is drawing the addressee's visual attention to a woman in the same room, the referring expression is exophoric. It is possible that, as well as attention being drawn to her physical presence, this particular woman will already have been referred to verbally. Since this may be so, "the line between exophoric and anaphoric reference is not always very sharp" and an instance of reference "may even be both at once" (Halliday & Hasan, 1976, p. 37).

Some commentators have applied value judgments on the basis of whether referent is established by reliance on contextual cues or on 'context free' statement. (Bernstein, 1971; Simon, 1981; Tough, 1976; Wells, 1981). Others, in contrast, concentrate solely on whether the listener has been provided with adequate information for comprehension, (Damico, 1985).

The interpretation of a referring expression may also rely on social context, the term general phoric being used. Social context is concerned with the experience, information and attitudes which can be assumed to be shared between conversational partners (or between individuals and their audiences). For example, if a speaker says "The pets were frightened", while talking about a household known also to the addressee(s), the referent will be established without these particular pets being physically present or having been previously mentioned. Grasping and taking into account what information and assumptions are shared between conversational partners, across a range of social and situational contexts, is a challenging task for young children.

A referent may thus be regarded as 'given' or mutually specific as the result of previous mention, as the result of its presence in the immediate physical setting or as the result of its specificity having been established in the shared experience or shared culture of the conversational partners. 'Given' referents may also, by a process known as entailment, confer specificity on entities which are associated with them. For example, someone can point to an empty bus and ask, "Where's the driver?" knowing that it will be assumed that the referent is the driver of that particular bus. Reference can equally appropriately be made to 'the driver', if a particular bus, though not physically present, is part of the current topic of conversation. A similar process comes into operation when the definite article is used in reference to a synonym for an established referent or to a term indicating a super-ordinate category of which it is a member. For example, once reference has been established to a particular dog, one can also use the terms 'the pooch' or 'that animal'. However, whether a term refers directly or indirectly to an entity which it already mutually specific, it cannot simply be used at random. It must be introduced at a point in the conversation where its link with the previously established referent is clear.

Intonation plays an important role in indicating whether a referent is to be regarded as newly introduced or as already in play in the conversation. A newly introduced item is indicated by tones that end in a falling movement (proclaiming tones); an item already in play is indicated by tones that end with a rising movement (referring tones). (Brazil, 1981; Wells, 1981). Thus the utterance "This dog is a puppet" can either indicate 'Talking about this dog, it is a puppet' (referring tone on 'dog') or 'Talking about puppets, this dog is one' (proclaiming tone on 'dog').


Those referring expressions which use pronominals, demonstratives (this, that, these, those, here, there, now, then) or the definite article typically involve co-reference, that is to say they indicate a referent identical to one established elsewhere in the discourse. Substitution and ellipsis, while also depending for their interpretation on a link being made with an established referent, typically involve co-classification. (Halliday & Hasan, 1985). When co-classification occurs, the established referent necessary to the interpretation of a referring expression is not identical to the current referent but is a member of an identical class. For example, a child, concerned to replace a hat which is too large, says, "Hey, I've got a little one in my bag". The child is clearly referring to a different hat from the one whose identity has already been established, but it is only by reference back to that previously specified hat that the referent "little one" can be understood.

Ellipsis differs from substitution in that instead of a term such as 'one' or 'so' being used, there is a 'substitution by zero', part of the structure of a sentence being omitted on the basis that its meaning is recoverable from the context. An instance of ellipsis would have occurred, if the child quoted above had been asked, "Who has a hat?" and had replied, "I have." Alternatively she could have said, "I have one" and this would also have been an example of ellipsis. This elliptical use of 'one' as an indefinite article (the plural form being 'some') needs to be distinguished from the use of 'one' as a substitute as in, "I'll show you which one I want" (where the plural form is 'ones'). This distinction is discussed by Halliday and Hasan (1976. p.159). It should be noted that substitution and ellipsis may refer back not only to individual referents but also to propositions.

A variant of ellipsis, known as sub-audibility (Perera; 1984, Quirk & Greenbaum; 1973) occurs when a speaker omits the beginning of a sentence and what is omitted can be inferred from the situation, rather than from elsewhere in the discourse. For example if a child holds up a puppet and says, "looks really good", the missing pronoun at the beginning of the utterance can be deduced from the situation.

Although co-referentiality is typically involved when pronominals, demonstratives and definite articles are used, and co-classification is, on the other hand, typical of substitution and ellipsis, both processes can occur in any of these five categories. Where a referring expression falls into the comparative category, as in the use of 'the same' or 'different', co-referentiality or co-classification are equally likely to be involved.


The referring expressions which come under this heading do not rely on co-referentiality or co-extension in order to be understood, but are concerned with the initial introduction of a referent into the discourse.

Unlike cohesive devices, these expressions are not completely reliant on context for their interpretation. Nevertheless, when a new referent is introduced context frequently plays a major role in its interpretation. The lexical item 'lamb', for example, will not establish the same referent if the physical location or the topic of the discourse is a butcher's shop as it will if the location or topic is the children's corner of a zoo. A further consideration is the attitudinal factor inherent in social context which influences, sometimes very subtly, the meaning of what is said. For example, an epithet such as 'rough' or 'old' may be understood as either complimentary or derogatory depending on the social context in which it is used, and to whom it is applied within that context. The acquisition of this aspect of meaning, which is crucial to one's functioning as a social being, is of pressing concern to young children.

Children need to be able to work out the intentions and expectations of their interactional partners, whether peers or powerful adults, with the complicating factor that these expectations may not always accord with surface features. (For example, what is formally a request for information, such as, "Could you stop making that noise?", may functionally constitute an order.) It has been argued that the need to interpret social expectations is more salient for young children than the need to analyse language for its formal meaning and that this explains some of the difficulties which they have with formal aspects of language. This effect is seen as being especially marked when children's communication skills are tested in a 'de-contextualised' experimental setting, in which making "human sense" of the situation is itself a problem for the child. (Donaldson, 1978). From this viewpoint, it is seen as important that observations of children's talk, and of its pragmatic aspects in particular, should be made under normal conversational conditions in which the variation in situational factors occurs naturally (Garvey, 1984).


An area which has been extensively researched in experimental settings, and which is directly relevant to the establishment of referent, is that of referential communication. Referential communication is concerned with those referring expressions which establish the identity of a referent by drawing attention to its identifying features. In a referential communication task a child is required either to give or to follow an instruction specifying the distinguishing features of a given item. For example, two children , with an opaque screen placed between them, are each given an identical set of pictures of a face. The pictures differ from each other in details such as the presence or absence of hair. One child has to identify a particular picture in the set on the basis of the other child's verbal description of it (Dickson, 1982). Children are thus required to analyse what attributes of a given referent distinguish it from the relevant alternatives and to convey this information to the partner. They are also required to grasp that this is the strategy necessary to perform the task. Reviews of studies of the referential communication task (Bowman, 1984; Bunce, 1991; Dickson, 1982) show that it presents considerable difficulties for young children and that, though performance improves steadily with age, full competence is not achieved until the age of about ten.

In the early studies of referential communication young children's difficulties were mainly attributed to a deficiency in 'role-taking' or presupposition: the ability to take the listener's perspective and tailor to this the information provided (Glucksberg, Krauss, & Weisberg, 1966). This approach was in keeping with the Piagetian view that children under the age of six or seven are 'egocentric', not having developed the capacity to shift from their own viewpoint to that of another person (Piaget, 1926, 1958). As evidence grew that young children often do adapt to the viewpoint of another (Donaldson, 1978; Lloyd & Beveridge, 1981; McTear, 1985b), emphasis in the referential communication studies was transferred to the difficulty children experience in analysing and coding the critical attributes of a referent. Of particular interest are a number of studies, summarised by Bowman (1984), which give evidence that pre-schoolers engaged in a task of this kind do not themselves grasp (without specific instruction) whether or not a description given to them is ambiguous. Since in this situation children are not required to take the viewpoint of another person, it follows that it is with the analysis of distinguishing features, and not with role-taking, that they are having difficulty.

There are naturally occurring situations in which a verbal exchange similar to that in a referential communication task could arise. For example, a child on the telephone (and therefore with all visual aspects of communication removed, as by the opaque screen) might ask a grandparent to buy a particular item from a set of similar toys. However, the tightly predetermined rules governing the task mean that it differs in a number of basic ways from the situation in which referent is established in naturally occurring conversation. While children may be confused by the lack of "human sense" involved in the artificial situation of a referential communication task, and do find the coding of critical attributes very difficult, it is also true that many of the demands involved in establishing referent in ordinary conversation are absent from the task. This is a consequence of the 'decontextualisation' which has been engineered in order to control the variables involved.

In most variations of the referential communication task, the competing referents which the child must analyse are confined to a visual array of slightly differing items, of which the partner is known to have a copy. This array constitutes the referential domain, "the range of objects from which the referent must be set apart." (Rommetveit, 1985, p. 186). In everyday conversation, however, it is far more often the case that the speaker has to deduce what the referential domain is by considering such matters as what the addressee is likely to know about objects of this type and what aspects of the referent (including its location) distinguish it from any item with which it could be confused. As in the referential communication task (Clark & Wilkes-Gibbs, 1986) the conversational partners may need to work together to check and clarify the identity of the referent. However in ordinary conversation the success of this cooperative enterprise will require the participants to distinguish the features of the referent not simply in relation to other items in a visual array, but in relation to shared experience and knowledge and to the context at that particular point in that particular conversation.

An important element in what constitutes context at any given point in a conversation is topic. Discourse topic can be defined simply as "what the participants are talking about" (Levinson, 1983, p.85). Schegloff has discussed the importance of topic analysis in establishing referent, pointing out that in interpreting referring expressions the assumption is made that they are relevant to "what is being talked to, what the focus is" (Schegloff, 1972., p.128). Referential communication tasks, however, "limit the topic of conversation in order to enable analysis of the child's message content" (Meline, 1988, p. 120). They thus eliminate the need for the child to participate in the mutual construction of topic and to situate the referent relevantly within this topic.

Referential communication tasks can have a role in therapy as an instructional game, which focuses children's attention on the careful monitoring of messages and on identifying the attributes of similar items and analysing how they differ. However, as we have seen, there are many dissimilarities between the demands of referential communication tasks and the demands involved in using identifying features to establish referent in everyday conversation. It is, therefore, not surprising to learn that a comprehensive review of the referential communication literature "revealed scant attention to... the validity of referential communication performance for assessing communicative competence in more natural settings." (Dickson, 1982, p. 134).


The meaning of individual lexical items is an important factor in establishing referent. Among the causes of difficulty in this area are gaps in the vocabulary of speaker or listener, differences in word meaning between dialectical or cultural groups, and the specific word finding difficulties associated with certain clinical conditions. Communication with young children must, of course, be geared to their level of semantic development. As already discussed, even when the referring expression consists of a single content word (such as 'lamb'), context (and in particular the current topic of the conversation) may be crucial in determining the identity of the referent. This is especially likely to be the case when the referring expression consists of a general word (such as 'person', 'thing', 'place', 'idea', or 'do'). Such words correspond to a whole class of lexical items (Halliday & Hasan, 1976). A similar situation occurs when the referring expression is an adverbial of time or place, such as 'afterwards' or 'outside', which needs to be related to context in order to be accurately understood. As Halliday and Hasan have pointed out, it is not possible to compile a definitive list of general words. Rather, there is a continuum from the most general to the most specific terms. As discussed in 1.4.4 above (in relation to the child who talked about Father Christmas getting ready to 'do' presents), what degree of specificity is appropriate will vary according to the context of the discourse and the speaker's intentions.


Information communicated by non-literal language is contrary to fact, yet intended to convey some truth or insight about the referent.

A referring expression is non-literal when expressed metaphorically. It is also non-literal when a speaker employs certain devices, such as irony and hyperbole, which involve giving false information with the intention that the listener will recognise it as false.

2.3.6.a Metaphor

A metaphor involves the juxtaposition of a topic and a vehicle, the topic being the subject of the metaphor and the vehicle being the means by which the speaker indirectly refers to the topic (Richards, 1936; Winner, 1988). For example, if a parent were to say to a small child, "You're my pussy cat", or "You're like a tornado today", 'you' in both cases would be the topic and 'pussy cat' and 'tornado' respectively would be the vehicle. The term 'ground' is used to refer to the attributes shared by topic and vehicle. The ground in the first of these examples is being small and pleasant to caress: in the second example the ground is being full of energy and destructiveness. Winner (1988) argues that the distinction between literal and non-literal (or metaphorical) similarity is dependent on whether there is a salience imbalance between vehicle and topic. Instances of literal similarity "involve matches between elements of the same conventional category that share properties highly salient to both" (p.98), as in, "An escalator is like a stairway." Instances of non-literal similarity "involve links between elements from different categories that share properties more salient to the vehicle than to the topic" (p. 98-99), as in, "An education is like a stairway." Whereas the function of a literal comparison is to elucidate the identity of the topic or referent, the function of metaphor is to throw new light on it (as in the notion of education being like a stairway in providing a means by which one can climb upwards). Using this distinction as a basis, the view is taken in the current study that where a metaphor is used there are three referents to be established, the topic, the vehicle and the shared ground. For example, in the statement, 'Laziness is like a stairway', although the referring expressions 'laziness' (topic) and 'stairway' (vehicle) establish referent, the ground of similarity between these two (which constitutes the third referent in the metaphor) remains unestablished. There would appear to be no attributes shared between a stairway and laziness and nothing is therefore added to the addressee's concept of laziness.

2.3.6.b Irony

The focus of metaphor is on an increased understanding of the referent. The focus of irony is on an increased understanding of the speaker's attitude and is communicated by means of a deliberate contrast between apparent and intended meaning. In stating the opposite of what both speaker and listener know to be the case (Winner, 1988), (or alternatively (Gibbs, 1989) in implying it), the speaker intends to convey an attitude towards the actual state of affairs. Most frequently this attitude is one of disapproval. For example saying, "It's a beautiful day!" when it is actually cold and wet, conveys an attitude of disapproval towards the weather. Similarly, to say "Thank you for waiting for me!" to someone who has actually forgotten to do so, conveys disapproval of the addressee's thoughtlessness. To interpret irony it is therefore necessary to infer not only whether or not speakers believe what they say but also whether or not they intend to deceive the listener. ("It's a beautiful day" could, after all, simply be a lie.)

Developmental aspects of metaphor and irony are discussed in 2.4.5



The developmental roots of the capacity to establish referent can be traced back to a very early stage in infancy, in the interactions between babies and care-givers. This early development culminates at a point, some time at the beginning of the second year of life, when children can be expected to show a good degree of competence at what has been called 'quasi-reference': the deliberate drawing of attention to a specific entity in the immediate environment (Bruner, 1983; Garvey, 1984; McTear, 1985b). Referring expressions, in spite of their increasing complexity as the child matures, retain this necessary basis in the establishment of a shared focus of attention. This shared focus is dependent on both attention getting and attention directing (McTear, 1984) and can be described as a state of intersubjectivity. Intersubjectivity has been defined as the situation attained when some aspect of the state of affairs at a given stage in a dyadic interaction "is brought into focus by one participant and jointly attended to by both of them" (Rommetveit, 1985, p.187).

By the age of two months babies maintain eye-contact with a caregiver, often to the accompaniment of vocalisations in which the beginnings of conversational turn-taking can be seen. By the age of about five months they will respond to an attention getting signal, such as the strongly intonated speaking of their name, by looking around them or directly towards the speaker. Care-givers usually help to shape the development of shared attention by dramatising their responses to the baby (for instance, by pulling an exaggerated face to encourage the baby to look at them) (Bruner, 1983). They will also invest infants' actions and vocalisations with an appropriate, but not necessarily intended, communicative function (Ninio & Bruner, 1976). For example, mothers have been observed to follow their babies' line of gaze and point to or name whatever the baby happens to be looking at (McTear, 1985b). In the last few months of their first year the babies themselves will follow a care-giver's shifting line of regard and by 12 months they start to transfer attention from the caregiver to the object which the caregiver is observing, then back to the caregiver, and back again to the object.

Babies appear to be just as determined as care-givers to initiate joint attention. Observation of children as young as four weeks suggests that they act to draw attention to themselves with, for example, a cry (McTear, 1985b). By about the age of five months the first form of request emerges when infants start to reach out towards specific items and a month or so later they start to give a vocal signal to request an object (Bruner, 1983). Somewhere around the end of the first year or the beginning of the second, pointing appears, serving the function not of requesting but, in Bruner's words, of "singling out the noteworthy" (p. 75). Typically, children first point, then look at the other person present, then point again, then look again, presumably to check that their companion is attending (Garvey, 1984). Children also frequently have a particular vocalisation such as 'oo' or 'da' which accompanies this pointing, and which is usually succeeded by a recognisable word such as 'look' or 'that'. At the same age they begin to point in response to being asked, "Where's the x?"

By around the time of their first birthday children are thus unmistakably involved in the joint enterprise of establishing referent. Although the current study is concerned with the establishment of referent between conversational partners, it should be pointed out that children may also be concerned to establish referent when speaking solely to themselves. Bruner (1983) describes a child, early in his second year, who had recently been watching rooks and magpies for the first time and, while sitting indoors by himself, pointed upwards and said "boe", his word for bird. There are many opportunities for young children to use referring expressions in a similar way during solitary talk and in running commentary on solitary make-believe play.


A distinction can be made between environmental topics, where the speaker draws attention to an entity in the immediate physical environment, and abstract topics, where the referent is not in the immediate environment (or at least not visible to the addressee(s)). Displaced reference is concerned with abstract topics and is often described as referring to the 'there and then' as opposed to the 'here and now'. During the second and third years of life there is a gradual emergence in children's conversations of talk about the 'there and then', an aspect of which is the use and understanding of referring expressions indicating referents which are not visible in the immediate physical setting (Weist, 1986).

Sachs (1983) recorded the emergence of displaced reference in the language of one child from the age of 11 months to 36 months, describing this in relation to the available literature on this aspect of development. At 17 months the child began regularly to name objects which were not in sight, and also to search for absent objects when they were named to her. At 22 months the function of comment on absent entities was added to that of request for these. Strategies other than pointing or touching were, of course, needed to establish the identities of these absent entities and where early referring expressions of this type were successful there was usually only one possible referent. This was achieved by the introduction of unique referents, such as 'Daddy' or 'the sun', and by the use of established conversational routines where the topic and structure of the conversation were familiar to the child and her adult partner. For example, this child and her parents had a routine conversation beginning, 'Where's the moon?' Ritual conversations of this kind, often involving the formula 'Where's the x?', are common between parents and very young children and their role in the growth of reference has been discussed by a number of observers (Bruner, 1983; Griffiths, 1986).

Reference to past events appears to emerge somewhat later than reference to absent objects. The child studied by Sachs began referring to the immediate past at the age of 22 months, to the earlier past at 26 months and to the future at 29 months. By 36 months she showed some competence in talking about shared past experiences and non-present objects. However, such topics were usually introduced by an adult and the great majority of the child's talk was about the here and now. An area of particular difficulty was that of providing identifying features. An exchange is quoted where the child, at the age of 2;8 wanted one of her parents to find "that book about dancing people" (p. 9) and became frantic and distressed when, in spite of the adult's attempts at clarification, the referent remained unestablished. Sachs comments on the important role played in conversation with children of this age by the adult's "guessing and/or probing," (p. 21) a point to be borne in mind when looking at the conversations of older children with developmental delay.

During the later pre-school years there is considerable development in all areas of pragmatic competence. Prutting and Kirchner (1987) state that "the developmental literature suggests that by age 5 children show some form (possibly not fully developed) of all thirty parameters evaluated on the Pragmatic Protocol" (p. 108). Johnston (1985) concludes that "by age five children... revise their own utterances when queried, ask for clarification of ambiguous requests, and specify the attributes of a referent" (p. 83). McTear (1984) summarises the achievements of the pre-school years as follows:

"Children develop the ability to get attention, identify discourse referents, taking into account their listener's state of knowledge, and secure an appropriate response" (p. 58).

Some aspects of the establishment of referent remain, however, to be fully acquired after the age of five, in particular those concerned with the use of cohesive devices to indicate referents which are already mutually specific. Children use some of these devices much less frequently than others. In an investigation of cohesion in the narratives of normal and language-disordered children aged 7;6 to 10;6, Liles (1985) found that the incidence of use of comparative reference, ellipsis and substitution was too low to warrant a statistical analysis. Developmental studies of cohesive devices have looked mainly at pronominal reference and at the use of determiners.


The competent use of pronouns is dependent on the ability to shift perspective and children's use of pronominal reference is thus of particular interest in relation to the ability to take the perspective of another person. The referent of personal pronouns, such as 'you', changes with a change of speaker. There is an acknowledgment of the difficulty this formalisation of shift in perspective can present to the very young child in the common habit among care-givers of referring to the child and themselves by name or title (e.g. "Baby show Daddy!") (Chiat, 1986). Loveland (1984) has reported a study which supported the view that children's acquisition of the I/you distinction coincides with their acquisition of the ability to distinguish between their own and other people's points of view on the concrete, spatial level (as shown, for example, in whether a child turns a toy bear around to face another person when asked to show that person the bear's nose). By the age of three, children can normally use and understand the distinction between first, second and third person, when the pronouns are used exophorically (Maratsos, 1979; Tanz, 1980). The contrasts of number, gender and case may, however, take up to an additional eighteen months to be acquired (Perera, 1984).

By the age of three, children have begun to use pronouns anaphorically, together with other proforms, including ellipsis (Garvey, 1984; McTear, 1984). The consolidation of this usage takes some years. Discussing the language of six to eight year olds, Piaget commented, "Pronouns...are used right and left without any indication of what they refer to. The other person is supposed to understand." (Piaget, 1926, p. 102). As with other aspects of development which were taken to be evidence of egocentricity, this view of children's use of pronominals has been very much modified in the intervening years. Self-corrections are one source of evidence that children as young as four are capable of consciously avoiding the use of pronouns which will not establish referent. Children of this age have been observed to correct a pronoun with the noun for which it is acting as substitute. In McTear's (1985b) longitudinal study of the conversations of his young daughter and her friend a number of examples are given of self-corrections of this type. For example, at 4;9 his daughter remarked, "<she> my friend Heather knows how to take it off herself" (p. 193). McTear describes these self-corrections as attempts to make an item more explicit for the listener.

The verbal disambiguation required to establish the referent of an anaphoric pronoun is, however, more difficult for young children than the non-verbal means (such as pointing) which can be used to disambiguate exophoric pronouns. In a study of conversations between eight pairs of pre-schoolers, ranging in age from 4;9 to 6;4, Van Hekken, Vergeer and Harris (1980) judged 64.2% of the anaphoric pronouns used by the children to be ambiguous in their establishment of referent. In contrast only 17.4% of those pronouns for which non-verbal disambiguation was available were judged to be ambiguous.

The difficulty which pre-schoolers experience with the referential function of anaphoric pronouns and other cohesive devices continues into the elementary school years. Discussing children's narrative skills, Liles (1985) cites studies which indicate that the actual use of cohesion is fairly stable by the age of six. However, cohesive adequacy, where "the information referred to by the cohesive tie is easily found and defined with no ambiguity" (p. 133) (in other words, where referent is established) is not stable at this age and its developmental progression has not yet been traced. Liles' work on narrative suggests that cohesive adequacy is fairly stable by the age of 7;6. However, Perera (1984) quotes data collected by Fawcett and Perkins (1980) which shows normal eight, and even twelve year olds, using pronouns which cannot be interpreted from the preceding context. Indeed even adults make unclear pronominal references (Maratsos, 1979).

An over-estimate of what is known by one's conversational partner, or a failure to adjust to the partner's physical viewpoint, are only two of the reasons that referring expressions may be inadequately specific. In the case of anaphoric pronouns other constraints include the requirement to select person, number and case as appropriate at that particular point in the discourse (for example by taking into account which character is speaking in a narrative), a grasp of the syntactic rules which determine when a pronoun can refer to a noun in the same sentence and when it cannot (Chomsky, 1969) and the need to take into consideration any competing referents to which the pronoun could equally well refer. Maratsos has commented that what is surprising is not that children have difficulties learning how to use pronouns and determiners, but "that they ever begin to get it right." (Maratsos, 1979, p. 239).


The contrast between the definite and indefinite article is used to some extent by three to four year olds to mark the distinction between new and given referents. However, the over-use of the definite article has been described as 'endemic' up to the age of at least seven years. (Karmiloff-Smith, 1979; Hickmann. 1986; Maratsos, 1979; Van Langendonck, 1984). Self-corrections of the definite to indefinite article begin to appear in the speech of five to seven years olds (Warden. 1976) and reach a high frequency in eight years olds, when the adult usage starts to be firmly established (Karmiloff-Smith, 1979; McTear, 1985b).

As we have seen, the referents which very young children introduce are all physically present and become 'given' once one's visual attention is drawn to them. In this situation, descriptive terms are used to give additional information, rather than to bear the brunt of identification, and it is appropriate to use the definite article on first mention of a referent. It is for the function of referring to entities that are already the focus of attention that children initially use 'the' (Karmiloff-Smith, 1979). At a somewhat later stage, they need to learn to recognise whether a referent will be unambiguously identifiable (and thus appropriately introduced by 'the' ) simply because it is in focus. For example, it is appropriate to say, "Give me the puppet", if there is only one puppet present at the time. but not if there are more than one.

The earliest use of the indefinite article by young children is to indicate the names of objects. The use of 'a' in a nominative sense ('it's a hat') is distinct from its use in an identifying sense ('there's a hat in my bag'). where it serves the function of introducing a particular item into the discourse. There is also a third function when 'a' is used in an indefinite sense ('I need a hat'). In informal situations an alternative form of indefinite reference is available. 'This', or 'these' can be used to indicate that a noun is specific for the speaker but not for the addressee (Perera, 1984).

In a detailed study of the acquisition of determiners by a group of French children, Karmiloff-Smith (1979) has argued that it is only when children have developed a meta-linguisitic grasp of the differing functions of "a" and "the" (at around the age of 8) that they achieve full competence in their use. Put in a situation which required the use of one linguistic form to express two different functions, many of the 5 to 8 year olds in her study used redundant markers, or slightly ungrammatical forms, to make a distinction between the two functions involved. For example, children of this age used 'la mienne de voiture' when picking out a referent from among a number of candidates and 'ma voiture' when giving additional information about a referent which was already the focus of attention. In contrast. from the age of about 8 they would simply use the term 'ma voiture' for both these functions, as an adult would.

The over-use of the definite article by young children therefore appears to be due in part to their difficulty in grasping that the indefinite article serves several functions, only one of which is the introduction of a specific, but previously unidentified, referent. The role of pluri-functionality here is an instance of an intra-linguistic, rather than a contextual factor influencing the acquisition of referring expressions.

Nevertheless. contextual factors are also an important influence on children's ability to make appropriate use of definite and indefinite referring expressions. The question of whether a referent can be taken as already 'given' from the point of view of the addressee is a complex one. A young child may assume that an addressee (in particular an authority figure) already knows of the existence of a referent which is familiar to the child (McTear, 1984) and may therefore refer to it as 'the'. The picture is complicated by evidence that in certain contexts adults may make a similar assumption. In a study of young children's use of definite and indefinite articles, Warden (1976) used a group of 20 year old students as controls. Contrary to expectation, the frequency with which these adults used the identifying expression 'a in describing pictures was unaffected by whether they were looking at the pictures with the experimenter or were telling him about them from the other side of the room. Warden comments that this unexpected finding is apparently due to some unknown contextual factor. A possible explanation might be that, since the experimenter had provided the pictures, the participants in the experiment assumed him to be familiar with them. If adults make such assumptions. it seems likely that children will do so too. It is also possible that the participants picture descriptions were influenced by classroom conventions of language use. Some educational material requires students to use the definite article in picture description. producing such sentences as 'The girl and the boy are playing with the dog'. A similar convention is used in many reading primers.

Problems in analysis can also arise when looking at informal conversation. One area of difficulty concerns reference to what Quirk and Greenbaum (1973) describe as 'things (which) are part of the cultural situation', (giving as an example somebody reading 'the paper' while 'the kettle' is boiling) (p. 73). It is not always easy to determine whether references of this kind are 'given' within the cultural context.


Young children's understanding of metaphor has been investigated in detail by Winner (1988), together with some discussion of their use of metaphor in spontaneous speech. Winner presents evidence that children start to use language metaphorically at the age of 2 to 3 years. In a description reminiscent of Bruner's (1983) characterisation of quasi-reference as "singling out the noteworthy" (p. 75), Winner (1988) describes the function of metaphorical language as being to point out "a resemblance that has struck [the child] as noteworthy" (p 92). The over-extensions typical of early language do not fall into this category since they are used by the child to fill a lexical gap and not to point out a noteworthy resemblance. For example, when a two year old, on first visiting a riding stable and seeing her elder sister on a horse, exclaims in alarm, "Lucy on big doggy!" (Neville, unpublished observation), 'doggy is used because the child has no word for horse. However, when another child calls a sewing-machine a 'table-horsey' on account of its shape (Winner, 1988), he is using a metaphor, since he knows this is not the actual name of the object.

The great majority of metaphors observed in pre-school children are either based on sensory resemblances. such as shape (as with the 'table-horsey') or occur during symbolic play when a child uses a pretend gesture to make one object stand for another. (For example, by putting an open book on a puppet's head and calling it a hat.) Children first make one object stand for another in this way at around the age of thirty months (Vygotsky, 1976) and doing so is often referred to as 'double knowledge'. Noteworthy resemblance between vehicle and topic is by no means always a feature of symbolic play metaphors, since on occasion children will make use of any handy object to stand in for a comb or a bed or whatever other item their play currently requires. In contrast, early metaphors which are solely verbal are likely to be firmly grounded in a sensory resemblance of some kind. (For example, spaghetti 'snakes' are long and thin.) Winner argues, however, that pre-school children are not aware of the distinction between literal and metaphorical resemblance. Even when children have reached an age where they do grasp that metaphor is distinguished by the novel attribution to the topic of salient aspects of the vehicle, the distinction between literal and metaphorical resemblance is not absolute but moves along a continuum. Metaphors themselves may gradually be transformed over time into literal terms (as in the expression 'a weak argument'), illustrating the difficulty which may exist in determining whether a term is literal or metaphorical.

A number of observational studies, discussed by Winner, show there to be a marked decline in the use of metaphor over the pre-school years and a marked preference in the conversation of children of elementary school age, especially between eight and ten, for literal rather than metaphorical similarities. Children of this age have been observed to actively reject metaphor, insisting, for example, that a colour cannot be loud. It appears that at this period, which Winner calls the literal or conventional stage, children are concerned to learn the conventions of their society and to distinguish between the factual and the imaginary.

The use of irony does not appear until children reach school age and Winner argues that this is because irony cannot be understood until children are able to reflect about beliefs. It is not until some time between the ages of six and nine years that children realise that a belief (whether their own or that of another) may be false. They do not therefore grasp that beliefs about the same thing may differ from one person to another, or in the same individual from one time to another. The irony of a remark such as "It's a beautiful day" cannot be established unless those involved can attribute to each other a belief about the conversational partner's belief state. (I need to grasp that you know I know the weather is not beautiful at all.) Deliberately false statements, of a less complex nature than those discussed by Winner, but like irony in not being intended to deceive the listener, do occur in the language of children under six. Instances of the use of hyperbole have been noted by the author while working with children aged four and five. For example, a boy aged 4:9 complained to a friend whose slowness was keeping both of them from going out to play, "Hurry up! We'll have to stay in here for a week!" (Neville unpublished observation.)

Much younger children have been observed to use humour derived from the incongruous labelling of objects. In a study of the origin and development of humour, McGhee (1979) describes a child of 22 months calling out, "Daddy, oggie-miaow!" and then

bursting into "somewhat encouraging, somewhat artificial laughter." (p. 65). Playfully false labelling of this kind is enjoyed by many two year olds (Garvey, 1984).

Deliberate deception of a conversational partner is said by Winner (1988) to first appear at about the age of five and the ability to infer deceptive intent in another to appear about a year later. In such cases the liar presumably intends the referent to be understood literally by the addressee and it does not therefore seem appropriate to categorise the referring expression as non-literal.


The preceding overview of the developmental factors involved in the establishment of referent has looked at the acquisition of the use of referring expressions to indicate entities in a number of areas: those in the immediate physical environment; those removed in time or place; those which are already mutually specific for the conversational partners and those which are non-literal. (No reference has been made to discourse deixis as the writer is not aware of any study which has examined its use by young children.) This acquisition process has been shown to take place gradually over the pre-school and early elementary years. to have its basis in interaction and to involve a complex combination of cognitive, linguistic and social skills. So complex a development can be expected to present problems for children with developmental language difficulties and some of these problems are considered in the following section.


Prutting and Kirchner's (1987) finding that Specificity/Accuracy was the parameter of the Pragmatic Protocol on which deficit occurred most frequently in a heterogeneous group of language disordered children was discussed in Chapter 1 (1.5.3). The relatively few studies which have been carried out across a population of children identified as having specifically pragmatic difficulties support the view that the establishment of referent is an area of major difficulty for such children (Bishop & Adams, 1989; Damico, 1985; Haines, 1985; Johnson, Johnston & Weinrich, 1984; Smedley, 1989).

The establishment of referent has been identified as a problem area in a number of individual case studies of children with pragmatic difficulties. Jones, Smedley and Jennings' (1986) case study of such a child, who was found to make inappropriate use of deictic and semantically vague terms. was discussed above (1.4.4). Damico describes a ten year old who scored in the normal range on tests of the structural aspects of language, such as the Clinical Evaluation of Language Functions (CELF) (Semel & Wiig, 1980), but whose conversation was reported by his teachers to be 'bizarre'. This child's "interactive partner was unsure of his meaning much of the time" (p. 186). McTear (1985b), in making a pragmatic analysis of the language of a child whose conversations were "bizarre and confused", found that the child "sometimes failed to take his listener's perspective into account and underspecified his information" (p. 245). Johnston (1985) looked at a conversation which had 'gone awry' between a mother and her 9 year old son (who had been diagnosed as language disordered with accompanying affective disturbance) and observed that most of the boy's formal problems "involve the mechanisms of reference specification" (p. 89).

What was observed in these studies of individuals is consistent with the findings of Smedley (1989) when reporting on work in the classroom over a period of years with children described as 'semantic-pragmatic language-disordered'. Smedley states that "most if not all semantic-pragmatic children...often fail to establish referents adequately" (p. 185). He goes on to make the important comment that understanding referents is only one among several aspects of pragmatics which require the use of inferencing both to extract information and to convey it accurately.

A finding which initially appears to run counter to those reported above was made by Adams and Bishop (1989). In their study a comparison was made between the conversations of children, aged 8 to 12, who fitted the clinical description of semantic pragmatic disorder, conversations of a group of language impaired children who did not fall into this category, and conversations of a group of controls, aged 4 to 12. Contrary to their expectations, Adams and Bishop found that the children in the semantic-pragmatic group were not distinguished from the children in the other groups by a more frequent use of unestablished referent, although there was a trend in this direction. The data on which this finding was based was, however, confined to pronominals and demonstratives. In a complementary study of the same set of conversations, Bishop and Adams (1989), instead of looking for occurrences of specific categories of conversational behaviour, took the approach of identifying any utterances which were judged to be inappropriate. When this approach was taken, the semantic-pragmatic group were characterised by high rates of inappropriateness in categories which included the following: the use of a wrong lexical item of the correct form class (e.g. "rinse it" (a fire) for "put it out") (p. 246); pseudo-ellipsis, the use of an elliptical form such as "I did" when no presupposed element was available (i.e. the addressee could not tell what it was the speaker did); and 'unestablished referent', here used to indicate the use of a pronoun or definite article for which there was no mutually specific referent. Analysed in this way the conversations of the children who had been assessed as having semantic-pragmatic disorder were indeed shown to be marked by a higher incidence of failure to establish referent.

The profile of inappropriateness which characterised the semantic-pragmatic group described by Bishop and Adams was not found in the other language impaired group in the study. This group, in fact, showed a lower percentage of inappropriate utterances on all parameters examined than did the control group. Bishop and Adams discuss one child who was an exception, showing a high level of conversational inappropriateness, although he had not been assessed as having a semantic-pragmatic disorder This child had phonological problems and Bishop and Adams speculate that this had disqualified him from being assessed as having the disorder, since one of its characteristics is said to be fluent. articulate speech. It may indeed be that a mis-match between pragmatic difficulties and good skills in other areas is an essential feature of semantic-pragmatic disorder'. Bishop and Adams make the point that a failure to convey precise meaning may be accepted as normal in a child with a limited command of language structures, but will create an impression of inappropriateness in a child who is using long and complex sentences.

That pragmatic impairment can occur in the absence of serious impairment of the more formal aspects of language supports the view of Friel-Patti and Conti-Ramsden (1984) that "structural aspects of language such as syntax and phonology may be based upon different sets of skills than communicative aspects of language such as pragmatics and discourse" (p. 172). Johnston (1985) takes this position further to suggest that "the discourse problem space" may present "a unique set of learning tasks" (p. 91). However, such views are speculative. We are currently, as Prutting and Kirchner (1987) have pointed out, in a period when it is necessary to gather the information on which to base a paradigm of the pragmatic aspects of language. A clear view of pragmatic impairment, and of the most effective therapy approaches for those affected by it, will only be possible on the basis of considerable further investigation and debate. In examining a major pragmatic parameter, the establishment of referent, in the conversations of a group of communicatively impaired children, this study aims to contribute to that investigation.


In the study described in 1.7.3, Prutting and Kirchner (1987) found that deficits in Specificity/Accuracy, or the establishment of referent, were more frequently responsible for an inappropriate communicative effect in the conversations of a heterogeneous group of language impaired children than were deficits on any other parameter of Prutting's Pragmatic Protocol. Liles (1985) also looked at failure to establish referent in a heterogeneous group of language impaired children, and found that their narratives contained a greater percentage of unestablished referents than did the narratives of an age-matched control group with normal language. In their study, however. Prutting and Kirchner (1987) noted a large range of variability within the group of language impaired children. They suggest this could reflect "the lack of homogeneity of subjects due to the general diagnostic classification used to select participants" (p. 52) and speculate that different subgroups of language impaired children, with distinct clinical profiles, might each be found to be associated with distinct patterns of pragmatic deficit.

A particular subgroup of language impaired children, those whose conversations are characterised by inappropriateness, is discussed in the studies described in 2.5 above. On the basis of these studies it can be hypothesised that there is a high frequency of the pragmatic deficit of unestablished referent in the conversations of this group of children, who will henceforth be referred to by the descriptive label 'communicatively impaired'.

The pilot study using a group design approach reported in Chapter 3 looks at the occurrence of unestablished referent in the conversations of five communicatively impaired children.

These children had been described on referral for therapy as being inappropriate in their conversational interaction. They had been assessed as within the normal range of non-verbal intelligence but as having impaired verbal comprehension. Impaired verbal comprehension in combination with relatively good abilities in the use of the structural aspects of language has been noted by a number of observers to be a feature of the sub-type of language disorder here termed communicative impairment (Adams & Bishop, 1987; Bishop & Rosenbloom, 1987; Coleman & Neville, 1984; Conti-Ramsden & Gunn, 1986; Rapin & Allen, 1985). Language Comprehension status rather than a spoken language age, such as reflected by mean length of utterance (MLU). was therefore considered to be appropriate as a matching variable in the current study (Schmitt & Meline, 1990)

Each of the communicatively impaired children was matched with one child for chronological age and with another for verbal comprehension level. It was hypothesised that in these triadic comparisons it would be found that unestablished referent was used more frequently by the communicatively impaired children than by the age-matched peers, but used at similar level of frequency by the communicatively impaired children and the younger children matched with them for verbal comprehension. Such a finding would support the further hypothesis that failure to establish referent is a feature of the language of communicatively impaired children and that it is associated with their delay in verbal comprehension.

It was argued in Chapter 1 (1.7.2) that in a study of this kind a dual approach is required. The aim of determining whether a particular pragmatic deficit is typical of a particular clinical population is extended, through the use of qualitative analysis, to an investigation of why this should be so. In Chapter 4 a detailed textual and interactional analysis is undertaken of the videoed and transcribed conversations which provide the data for the current study.

The aim of this qualitative analysis is not only to investigate how the instances of unestablished referent in the conversations arise, but also to reach an understanding of what constitutes the inappropriate communicative effect with which they are associated. Taking the pragmatics-as-cause-effect approach, this communicative effect is seen as jointly constructed by the conversational partners. It may be, as Bishop and Adams (1989) suggest, that a mismatch between a child s failure to convey precise meaning and relatively good skills in the structural aspects of language creates an impression of inappropriateness for the addressee, who then reacts to the communicatively impaired child as 'odd . Alternatively, there may be features of the communicatively impaired children's failure to establish referent which are not found in the language of normally developing children at a similar level of verbal comprehension, and which are responsible (at least in part) for the inappropriate communicative effect which occurs.

Using Prutting's Pragmatic Protocol as a model, an organisational framework, the Protocol for the Establishment of Referent (P.E.R.), was developed by the author for use in the qualitative analysis carried out in the current study, and as a potential clinical tool.

In summary, the aim of Chapter 3 is to investigate the frequency with which unestablished referents are used by the communicatively impaired children, in comparison with their age matched and with their verbal comprehension matched peers. Chapter 4 aims to determine how these instances of unestablished referent arise and what constitutes the inappropriateness of their communicative effect.

The findings of this investigation are of considerable relevance to the practical issue of how best to undertake therapy with communicatively impaired children. This issue is discussed in the concluding section of Chapter 4.

Go to Chapter 3
Return to Contents


Hosted by