Discourse Processing in Attention-Deficit ... - Semantic Scholar

2 downloads 0 Views 362KB Size Report
Jun 4, 2008 - 1969), where a boy attempts to find his pet frog that has escaped from its .... We expected significantly fewer correct uses of the perfect tenses in ...
J Log Lang Inf (2008) 17:467–487 DOI 10.1007/s10849-008-9066-5

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD) Michiel van Lambalgen · Claudia van Kruistum · Esther Parigger

Published online: 4 June 2008 © Springer Science+Business Media B.V. 2008

Abstract ADHD is a psychiatric disorder characterised by persistent and developmentally inappropriate levels of inattention, impulsivity and hyperactivity. It is known that children with ADHD tend to produce incoherent discourses, e.g. by narrating events out of sequence. Here the aetiology of ADHD becomes of interest. One prominent theory is that ADHD is an executive function disorder, showing deficiencies of planning. Given the close link between planning, verb tense and discourse coherence postulated in van Lambalgen and Hamm (The proper treatment of events, 2004), we predicted specific deviations in the verb tenses produced by children with ADHD. Here we report on an experiment corroborating these predictions. Keywords

ADHD · Verb tense · Planning · Discourse models

In a pioneering study, Trabasso and Stein (1994) have argued that discourse production, for instance in a narrative task, requires a form of planning. In van Lambalgen and Hamm (2004), van Lambalgen and Hamm have shown how some of the mechanisms with which discourse coherence can be achieved—tense, aspect, adverbials—can also

This article is based on the theoretical analysis of executive function given by Stenning and van Lambalgen (2007). The analysis of the narratives of the children with ADHD is due to the second and third authors. The application to tense processing owes much to conversations with Giosue Baggio. M. van Lambalgen (B) · C. van Kruistum Institute for Logic, Language and Computation, University of Amsterdem, Nieuwe Doelenstraat 15, 1012CP Amsterdam, The Netherlands e-mail: [email protected] E. Parigger Amsterdam Centre for Language and Communication, University of Amsterdam, Spuistraat 210, 1012VT Amsterdam, The Netherlands

123

468

Michiel van Lambalgen et al.

be viewed as instances of planning. It is therefore interesting to pursue the hypothesis that developmental orders in which planning is known to be compromised lead to deviant forms of discourse production. For this purpose it turns out to be useful to consider planning in the wider context of executive function and its disorders. ‘Executive function’ is an umbrella term for processes responsible for higher-level action monitoring and control that are necessary for maintaining a goal, and achieving it in possibly adverse circumstances. Several psychiatric disorders are accompanied by what seems to be executive dysfunction, and in fact executive failures have been hypothesised to cause these disorders. For instance, in ADHD one can observe difficulties with response inhibition, vigilance, working memory and planning (for a recent overview see Willcutt et al. (2005)), and it is tempting to suppose that these problems are the cause of the behavioural problems associated with ADHD. The problem with such hypotheses is that executive function is not a very well-defined concept (and neither is ADHD, for that matter). The concept of executive function originated from neuropsychological study of patients with impairments in this area, and the field still suffers from lack of a comprehensive positive characterisation of just what executive function is. There is also no unanimity on how to partition executive function into meaningful sub-components. This is unfortunate if one’s aim is to try to view different psychiatric disorders as caused by failures in different sets of components of executive function, thus moving beyond the general statement that executive function is compromised in psychiatric disorders. In Stenning and van Lambalgen (2007) it is argued that logical modelling is useful in providing executive function with conceptual coherence. It is also shown that logical modelling leads to testable predictions in the behavioural domain. The example considered there is autism; the prediction is that autists behave very differently from normals in the so-called ‘suppression task’, a reasoning task which requires the ability to withdraw an earlier conclusion in the light of new information. In the meantime this prediction has been verified (Pijnacker et al. 2007). In the present paper we apply the approach to ADHD, viewed tentatively as executive dysfunction. We take our cue from data showing that narratives of children with ADHD exhibit (i) retelling of story events out of sequence, and (ii) ambiguous anaphoric references to events (Purvis and Tannock 1997). This indicates that in these children temporal semantics is compromised. In van Lambalgen and Hamm (2004) it was shown that temporal semantics can be viewed as mediated by the human planning system, hence by executive function, and from this we derive predictions concerning the use of tense, aspect and temporal adverbials in narratives produced by children with ADHD. The predictions have been tested by the second author and the results are reported here.

1 A Logical Description of Executive Function Taking executive function to be composed of planning, initiation, inhibition, monitoring, co-ordination and control of action sequences, leading toward a goal held in working memory, Stenning and van Lambalgen (2007) claimed that the operation of executive function can be described as a form of conditional reasoning,

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

469

and aberrations thereof. The authors identified closed world reasoning as a central component—corresponding to inhibition—of the logical description of executive function, and showed, both theoretically and experimentally, that executive dysfunction can be viewed as a failure in closed world reasoning. It may seem paradoxical that a largely automatic process such as executive function is associated with reasoning, which seems to involve conscious deliberation. But it is the association of reasoning with conscious deliberation that is wrong: whether or not reasoning needs consciousness depends very much on the computational complexity of the type of reasoning that the reasoner engages in. Classical reasoning is notoriously complex, but closed world reasoning is not. For our present purposes—relating planning to temporal semantics and deviant use of tense—we need not go deeply into closed world reasoning, but we must say something about the formal analogues of planning and working memory, and the role of goals therein. Planning consists in the construction of a sequence of actions which will achieve a given goal, taking into account properties of the world and the agent, and also events that might occur in the world. The relevant properties include stable causal relationships obtaining in the world, which can be formulated as suitable conditionals. To express these relationships in a fully precise manner one needs a formalism such as the event calculus (see for instance van Lambalgen and Hamm (2004), but for purposes of exposition we use propositional logic here. When we come to deviations of verb tenses in ADHD we will use a richer formalism. From a logical perspective, planning can be viewed as an instance of logic programming.1 A logic program is a finite database of conditionals of the form ϕ → q, where q is atomic. A goal is a finite list of literals,2 written ?L 1 , . . . , L n . In the context of planning, the goal represents a description of the desired state. The purpose of the computation is to determine what conditions must be met to satisfy the goal, using only the conditionals in the database. A computation starting from a query ? A with a database containing the conditionals B1 , . . . , Bn → A and C1 , . . . , Cm → Bi can be pictured as a tree as in Fig. 1. If the computation ends with the empty goal, this means that the original goal can be satisfied. If the computation terminates, but the last goal is non-empty, this means that the database must be updated with literals in the last goal in order to make the original goal satisfiable. In the case of planning these updates will consist of reports that the actions leading toward the goal have been executed.3 We view the maintenance of the goal stack—the currently active goal together with the goals from which it was derived—as as a function of working memory. Planning itself involves more, in particular search for relevant conditionals in declarative memory, and unification of the currently active goal with the consequent of the conditional selected.

1 The classical reference is Kowalski (1979); a modern exposition can be found in Doets (1994). 2 A literal is either an atom or the negation of an atom. 3 The above description of a computation in logic programming uses goals as integrity constraints, as developed by Kowalski (1995). The application of integrity constraints to verb tenses used in the next section is developed in van Lambalgen and Hamm (2004), Chap. 8.

123

470

Michiel van Lambalgen et al.

Fig. 1 An example of a computation

2 Verb Tenses from Planning We now connect the preceding abstract considerations to the domain we are interested in, verb tenses. In the book The proper treatment of events (van Lambalgen and Hamm 2004) the main function of tense and aspect is considered to be the construction of event structures from discourses. This construction is viewed as a planning problem: the same mechanism which in planning constructs a sequence of actions is responsible for the construction of event structures from discourses. In this set-up, verb tenses are represented as goals in the same sense as goals are used in planning. More precisely, the task of the speaker is to formulate a goal (by means of a sentence in natural language) which allows the hearer to update his or her event structure (‘discourse model’) with the new event expressed by the sentence. The event structure comprises a (partial) order of the events in terms of relations such as ‘precedes’ and ‘overlaps’, and rudimentary indications of the durations of, and spacing between events. Ideally, such a goal has two components: (1) location of the event in past, present or future (2) meshing the event with events introduced previously Thus, the computational function of verb tenses is much more than locating an event with respect to now; the real computational burden is borne by the incorporation of the event in the discourse. For our purposes, the interesting question is how this burden is divided over speaker and hearer. A ‘pragmatically sensitive’ or ‘emphatic’ speaker will formulate the goal in such a way that the computational burden on the hearer is minimal. By contrast, a speaker with no concern for the hearer, may leave out the contextual information that would facilitate the hearer’s computations. An example will make this clearer. Suppose a speaker wants to convey the message that John pushed Max, and that as a consequence Max fell. One way to do this is via the mini-discourse Max fell. John pushed him. Note that here the order of the sentences is the reverse of the order of events. The goals conveyed by this discourse are (1) ‘update discourse with past event e1 = fell(m) and fit e1 in context’ (2) ‘update discourse with past event e2 = push(j,m) and fit e2 in context’

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

471

The first part of these goals can be executed immediately; one does not have to consider a reduction to other goals. It is the fitting into the context that is computationally costly; why this is so can be seen if the condition ‘fit e2 in context’ is spelled out. For a start, the general form of the condition must be something like (1) fit e2 in context {e, e , e , . . .}, where {e, e , e , . . .} is a set of variables for events. These variables have to be unified with the events activated in the given context, which will come with an event ordering expressible in terms of < (‘precedes’).4 The instruction ‘fit’ then asks for an extension of this ordering with e2 . In the case at hand, the general condition reduces to ‘fit e2 in context {e1 }’, and the instruction ‘fit’ can be viewed as asking which if any of the queries ?e1 < e2 and ?e2 < e1 can be made to succeed. It is hypothesised here that it is the planning system which determines the event structure from the discourse. In order to do so, the planning system recruits causal knowledge as well as the principle that causes precede effects. Applied to the case at hand, the planning system scans declarative memory for causal connections between e1 and e2 and finds (roughly) ‘e2 is a cause of e1 ’. This fixes the temporal order of e1 and e2 , with e2 preceding e1 .5 Viewed in this manner, the computation for the hearer involves (1) keeping the context active in working memory, that is, events and their order (2) search and unification prompted by the ‘fitting’ condition (3) execute in parallel the backtracking derivations that determine the place of the new event with respect to the old event order (4) this step in itself involves a search for applicable causal information, and unification This computation is more complex than the simple location of the event in past, present, or future, which can be read off immediately from the verb tense. If the preceding is a good, albeit high-level, description of the computations involved in producing and comprehending verb tenses, we can identify factors which affect the computational load for speaker and/or hearer. (i) The speaker must produce a goal that is satisfiable, so that the hearer can start a successful computation from the given goal. An example6 where this is not the case is (2) Hij viel bovenop het gewei. Het hert is kwaad en liep naar de afgrond. He fell onto the antlers. The deer is angry and walked to the ravine. [9yrs, ADHD] 4 The event order need not be total; indeed it may be the function of the sentence under consideration to fill in missing bits. 5 Inferring this event order is a defeasible process; if the discourse had continued beyond ‘him’, as in

Max fell. John pushed him, or rather what was left of him, over the edge. the order of e1 and e2 is now different, since it conjures up a scenario like the following: John does something particularly nasty and bloody to Max which makes him fall, near the edge of precipice; he then shoves the body over the edge of the precipice: thus e1 precedes e2 . 6 Taken from the experiment reported in Sect. 4. The utterances are labelled with the age of the child and

the group to which it belongs, ADHD or C(ontrol)G(roup).

123

472

Michiel van Lambalgen et al.

The past tensed ‘walked’ is incongruous after the present tensed ‘is’. The past tense takes its reference point (which must lie in the past) from the context, but the first conjunct has identified that reference point with the utterance time.7 Note that the transition from ‘fell’ in the first sentence to ‘is’ in the second can still be interpreted (with some effort) as a transition to the narrative present, which formally means making the ‘now’ representing utterance time a variable that can be unified with the preceding reference point. No such reinterpretation is possible however for the transition from ‘is’ to ‘walked’ within the second sentence. It is then an interesting question what is entailed by the requirement that the speaker formulates a satisfiable goal. One may at first be tempted to think this is a matter of morpho-syntax only, not semantics. But a uniform morpho-syntactic consistency requirement like ‘use the same tense through the whole discourse’ quickly falls foul of more creative language use such as the narrative present. This means that the consistency requirement must be of a semantic nature, so that the speaker must in essence perform the same computation as the hearer in order to see whether the goal he is in the process of formulating is satisfiable. As a consequence, the speaker can ease his own computational burden by omitting the satisfiability check, or by leaving out tense in indicative sentences at all, so that, if the model of (Van Lambalgen and Hamm 2004) is correct, there is no goal whose satisfiability is to be checked in the first place. As we shall see in greater detail later, some children with ADHD indeed adopt this strategy, by producing utterances without any verb8 , or by resorting to direct speech.9 (ii) The speaker eases the computational burden of the hearer if the goal (i.e. the sentence produced) contains sufficiently many context setting elements. For upon hearing a particular tense form, say a past tense, the hearer activates a goal of the form (1), and he has to unify the variables in the goal with contextual material. If these are provided by the speaker, this eliminates a search step. Context-setting can be achieved by temporal adverbials, or by connecting clauses such as ‘when’, ‘and then’, ‘because’ or ‘but’. The following examples illustrate the contrast. (3) a. Max fell. John pushed him. b. Max fell because John pushed him. A defining condition for ‘e1 because e2 ’ is that e2 precedes e1 , so that the computation of the event order in the second sentence is trivial as compared to that for the first sentence. 7 Thus there is no unification which makes this goal satisfiable. This impossibility of unification leaves interesting traces in the EEG. See Baggio (to appear) for discussion. 8 E.g.,

(i) “Oeh.” “Boem” “Stoute hond…stoute hond” “Ooh” “Boom” Bad dog…bad dog” [9yrs, ADHD] 9 That is, by reporting what they imagine the protagonists to have said. Direct speech can take the form of indicatives, of questions or of imperatives. The tenses of the verbs in direct speech, even in indicatives, do not contribute to the construction of the discourse model, so do not need a consistency check.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

473

(iii) The choice of tense also influences the complexity of production and comprehension of the goal. In a narrative task, a speaker not using direct speech has a choice between past tense and present tense, where the latter must be understood as narrative present. In the present tense, utterance time, reference time and event time coincide. Therefore in the real use of the present tense no context-setting elements are necessary. This is different if the interpretation is that of a narrative present. Here are two examples from Google: (4) a. So, on Tuesday I arrive at work and my laptop fails to start up. It powers on and then shuts down 2–3 seconds later. b. I caught a train early from Frankfurt to get to Kassel in time for a press conference on Tuesday. I arrive well on time. Everything is much slower here, I am glad, I need to pace myself a little. In (4-a), the adverbial ‘on Tuesday’ must refer to a past Tuesday, and still the main clause uses the present tense. In (4-b) the context is set by an event which is explicitly in the past, but nonetheless the remaining sentences of this discourse are in the present tense. This means that the ‘now’ of the present tense in its narrative use actually becomes a variable that must be unified with already established reference points; it is no longer a constant with fixed reference. If the goal provided by the speaker does not include reference points to which ‘now’ can be anchored, the computational burden on the hearer is increased. Thus, a sequence of clauses in present tense without temporal adverbials or other context-setting elements may need considerable computation on the part of the hearer to determine the ‘spacing’ of events. 3 ADHD, Executive Function and the Art of Telling a Story ADHD is characterised by persistent and developmentally inappropriate levels of inattention, impulsivity and hyperactivity, and about 2% of children (mainly boys) are severely affected; 3–6% suffer from less severe symptoms.10 It has been hypothesised to be an executive function disorder, and indeed children with ADHD score significantly lower on a number of standard tests measuring components of executive function, such as planning, inhibition, and self-monitoring. The precise pattern of executive deficits in ADHD is not yet known, and it is not yet determined whether there is a single executive deficit that explains most of the symptoms. Below we will investigate consequences of the hypothesis that goal maintenance is affected in ADHD, evidence for which can be found in Shue and Douglas (1992), Pennington and Ozonoff (1996). We will be particularly interested in the effect of deficient goal maintenance on language processing, and we therefore briefly introduce some of the relevant neuropsychological data.

10 These are figures for the Netherlands, supplied by the Gezondheidsraad.

123

474

Michiel van Lambalgen et al.

3.1 Executive Deficits in ADHD Shue and Douglas (1992) have shown that children with ADHD score significantly worse than neurotypical subjects on a family of tasks which involve synthesis and execution of a plan achieving a given goal. The paradigm case is the ‘Go/NoGo’ task whose basic structure is as follows: on one type of stimulus the subject has to press a key (this is the ‘go’ stimulus), on another type of stimulus the subject must not do anything (the ‘no go’ stimulus). Stimuli are presented in random order. In the version of Shue and Douglas (1992), the ‘go’ stimulus is a card showing an apple, and the ‘no go’ stimulus a card showing an ice cream;11 the response is pressing a key. Viewed as a (rather minimal) planning problem the computation goes like this. The goal is ‘do(x) now’, with x a variable to be instantiated. There are two rules, which can be formalised as ‘if apple, do(go)’ and ‘if ice cream, do(no go)’. This means there are two possible unifications for the variable x in the given goal, the actions go and no go, which reduce the goal to satisfying one of the preconditions, apple or ice cream. The (unique) successful unification then determines the action to be performed. The plan P synthesised by this computation is of the form IF C1 THEN A1 & IF C2 THEN ¬A1 Of course if all goes well, after a few trials performance is determined by the automatically executed plan P (which proceeds forwards from stimulus to action), and not by the explicit computation of this plan (which proceeds backwards from goal to stimulus). But performance on this task in children with ADHD is significantly impaired in that many ‘no go’ trials lead to a ‘go’ response, and this suggests that the initial computation (unification and reduction of the goal ‘do(x) now’) is not executed correctly.12 If the ‘go’ response is conceived as an unconditional response, that is, if the goal is simplified to ‘do(go)’, it does not require any computation. We thus view the performance of children with ADHD on these tasks as evidence for the difficulty of keeping a complex goal active in working memory, and the tendency to simplify that goal as a consequence. It is this interpretation13 that we shall apply to the analysis of discourse production, in particular verb tenses, below. The link from ‘Go/NoGo’ tasks to verb tenses is provided by the analysis of verb tenses in terms of complex goals maintained in working memory, as explained in Sect. 2. Indeed, the apparently problematic ingredients in the ‘Go/NoGo’ task: maintaining the goal, unification and then reduction of the goal, are very prominent in the computation of verb tenses.

11 There are 20 cards for each of the ‘go’ and ‘no go’ stimuli, so it is not a matter of building up a response bias. 12 Unfortunately, it is not reported in Shue and Douglas (1992) whether there are symmetrical errors with

the ‘go’ trials. 13 We are aware that the results on the Go/NoGo task are often interpreted as evidence of a failure in

response inhibition. However, such a gloss on the results makes it difficult to understand why autists, known to have difficulty with ‘inhibiting the prepotent response’ (Russell 2002), do well on this particular Go/NoGo task (Ozonoff et al. 1994). This interpretational difficulty shows, if nothing else, how important it is to come up with precise formal analyses of the tasks involved.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

475

Fig. 2 Pictures 1–3 from “Frog, where are you?”

3.2 Discourse Production in ADHD Language ability also seems to be affected in ADHD, in particular with regard to semantics and pragmatics. This has been investigated using story telling tasks, in two forms. In Purvis and Tannock (1997) a folk tale ‘The father, his son and their donkey’ was read to the children, who then had to repeat the story. In Blankenstijn and Scheper (2003) the ‘Frog story’ paradigm was used, in which children were asked to narrate a sequence of 24 scenes in a picture storybook called Frog, where are you? (Mayer 1969), where a boy attempts to find his pet frog that has escaped from its jar. This is a classic experimental paradigm for investigating the acquisition of temporal notions in children. See Berman and Slobin (1994) for methods, results, and last but not least, the frog pictures themselves; the opening scenes are given in Fig. 2. The drawings depict various failed search attempts, until the boy finds his frog by accident. The purpose of the experiment is to investigate what linguistic devices, in particular temporal expressions, children use to narrate the story, as a function of age and mental condition. The narrative tasks in Purvis and Tannock (1997) and Russell (2002) both seem to involve executive function, in particular planning and goal maintenance. One general connection between narration and planning has been illuminated in Trabasso and Stein (1994), whose title says it all: “Using goal-plan knowledge to merge the past

123

476

Michiel van Lambalgen et al.

with the present and future in narrating events on line”. The main idea is that the events depicted in the book are naturally structured in time as a sequence of actions aimed toward achieving a goal, and hence the narration is like the unfolding of a plan. Executive difficulties are therefore likely to result in deviant narration patterns, and indeed several such have been observed. For instance Purvis and Tannock (Purvis and Tannock, 1997, pp. 136) observed a tendency to retell story events out of sequence; this reflects a breakdown in overall goal–plan organisation. Another phenomenon that these authors noticed, ambiguous anaphoric references to events, will be of special interest of us, since it reflects a more local breakdown of planning, a failure to achieve discourse cohesion. In particular we will see that children with ADHD have various difficulties with using verb tenses, thus causing breakdown of discourse cohesion. 3.3 Deviant verb Tenses and ADHD We are now in a position to formulate hypotheses on the the production and comprehension of verb tenses in narrative tasks by children with ADHD. For production one would generally expect decreased coherence of the narrative produced. That is, a speaker with ADHD will attempt to simplify the goal conveyed by his utterance even if that means increasing the processing load of the hearer. Based on a coarser form of the analysis given in Sect. 2, Stenning and van Lambalgen (2007) gave the following list of phenomena they expected to see. 3.3.1 Anchoring To simplify a goal, the child with ADHD may be tempted to omit temporal anchors where they are obligatory, as in the simple past tense (Steedman, 1997, pp. 906). Out of the blue a sentence like (5-a) is infelicitous, but with a temporal adverbial or a subordinate clause, felicity is restored, as in (5-b). (5)

a. It rained. b. Yesterday/When I stepped outside, it rained.

Now suppose the speaker is a child with ADHD. Producing a sentence like (5-b) requires the explicit integration of context and main event.14 If in the mind of the child with ADHD the goal corresponding to the past tense is simplified by dropping the meshing part, this should show as an increased production of ‘bare’ past tenses as in (5-a). 3.3.2 Perfect Tenses A more subtle deviation is predicted to occur with the production of perfect tenses. Generally, the goal corresponding to a perfect tense has the following structure (1) locate the event in the past of the reference time 14 See for example (van Lambalgen and Hamm, 2004, pp. 112–114) for a discussion of when clauses which

establishes this point.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

477

(2) the reference time is situated in the state resulting from the event (3) for the past perfect, the reference time must be determined by context Unlike the case of the past tense, here the first instruction cannot be executed without the unification demanded by the third instruction. This is definitely a problem in the case of the past perfect, since there the reference time is earlier than the utterance time. The problem also occurs for the present perfect when it is used in the context of the narrative present, since the story—‘now’ is different from the utterance time. Furthermore, the use of a perfect tense in an utterance constrains the use of a simple tense in the next utterance, since the event must be located at the reference time given by the perfect. The use of the perfects requires some foresight and planning as well as integration with preceding material. Here is an example of what can go wrong. A scene in the beginning of the ‘Frog story’ (Fig. 2) could be described thus (6) Toen ging het jongetje slapen. En toen wou de kikker eruit. Want hij had er geen dop op gedaan. Then the boy went to sleep. And then the frog wanted to get out. Because he hadn’t put a cap on [the jar]. [8yrs, CG] but also as (7) Een jongetje had een kikker gevangen of, ja. En hij heeft d’r in een potje gedaan. En toen ’s avonds liep de kikker eruit. En toen was het dag. En toen werd het kind wakker. A boy had caught a frog, or something like that, yes. And he has put her in a jar. And then at night the frog walked out. And then daytime came. And the child woke up. [9yrs, ADHD] The use of the past perfect in (6) is fine: the state resulting from not putting a cap on the jar is thereby shown to hold at the time of the frog’s escape. The use of the present perfect in the second sentence of (7) is infelicitous in combination with the simple past tense of the third sentence. The dominant tense of this passage is the past tense, and the use of the present perfect leads to inconsistency. We expected significantly fewer correct uses of the perfect tenses in children with ADHD. The present population (cf. Sect. 4) with children aged 7–9, is not ideal for testing this prediction however, since also normal nine-year olds use the perfects much less than adults. 3.3.3 General form of the Predictions The most general prediction that follows from the analysis given in Sect. 2is: (8) A speaker with ADHD simplifies the goals corresponding to tenses at the expense of the hearer. We list here a number of ways in which these goals can be simplified. An extreme form of simplification is not to use tensed verbs at all. As explained in footnote 9, another form of simplification is to use direct speech, with or without tensed verbs. Together with the complexity-reducing strategies given in Sects. 3.3.1 and 3.3.2, this gives us

123

478

Michiel van Lambalgen et al.

at least four strategies. The theory identifies other strategies as well: one would for instance expect fewer utterances in which several events are connected. The prediction (8) should then lead to the following observable consequence (9) Children with ADHD use a complexity-reducing strategy significantly more often than control children. 4 A Narrative Study 4.1 Subjects In the choice of subjects, age is the determining factor. Deviations are more likely to show at an age where the expression of temporality is still in development. Fortunately, we have the results of Berman and Slobin (1994) to build upon. They found that children as young as 3–4 years of age are able to produce pieces of discourse which are both intelligible and relevant to the task at hand (Berman and Slobin, 1994, pp. 42). However, [t]he ability to produce extended narratives which are well-formed in terms of hierarchical thematic structuring and global organisation emerges relatively late, well beyond the period when children produce sentences which are syntactically impeccable (Berman and Slobin 1994, pp. 42–43). As against this early emergence of structural proficiency, there are some forms in each language which do not occur at all in the texts of the younger children, being confined to occasional appearances among the 9-year-olds, or to strictly adult usage (Berman and Slobin 1994, pp. 599). At 3, 4 or 5 years of age, children still have a long way to go, even though their texts show clear signs of temporal organisation (Berman and Slobin, 1994, pp. 55). The past perfect (‘had read’, ‘had given’) is an example Berman and Slobin give of a form that is acquired relatively late (Berman and Slobin, 1994, pp. 599). In addition, only in the narrations of the 9-year-olds explicit reference is made to causal and not merely temporally sequential connections between series of events (Berman and Slobin, 1994, pp. 70), and even then this happens only occasionally. Their narratives still are far from mature, but a significant development seems to take place across the 5–9 age range. For practical reasons we had to restrict ourselves to the age range 7–9. This means that there is much room for improvement of the present study by extending the age range on both sides. As will be seen below, some of the phenomena we are interested in are only present in rudimentary form in normal 9 year olds. The group of children with ADHD comprised 26 subjects, the control group 34 subjects.15 The children with ADHD all fell in the so-called ‘combined type’ category, meaning they have both attention and hyperactivity problems. The breakdown according to age group is given in Table 1. An awkward feature of our ADHD population is the presence of comorbidity: 22 out of 26 children have other disorders as well. 15 The transcripts come from three sources: Roelofs (1998), van der Meulen (2003) en Aarssen (1996).

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD) Table 1 Number of subjects in different age groups

479

Age (years)

Control group (Boys)

ADHD group (Boys)

7 8 9 Total

6 (5) 11 (7) 17 (9) 34 (21)

5 (5) 8 (8) 13 (10) 26 (23)

Usually, the child shows one other disorder in addition to the primary one (38.5%), but there one can also find two (11.5%), three (26.9%) or even more non-primary disorders. Examples of such disorders are behavioural disorders (16.6% of all comorbid disorders), anxiety disorders (39.6%) and not being toilet trained (25%). These figures unfortunately make it unlikely that all children in the ADHD group have the same executive function profile. Almost all ADHD children used medication (92.3%), mostly Ritalin. However, the children discontinued their use of Ritalin prior to the day of testing, and due to its short decay time the medication could not influence performance. 4.2 Materials The children narrated the storybook Frog, where are you? (Mayer 1969), consisting of 24 pictures. They were allowed to go through the book first. They then narrated the story to go with the pictures while leafing through the book again. The researcher could not see the pictures and emphasised: I dont know the story, it is your story.Ó Note that in the study of Berman and Slobin, the story was always told to a listener who already knew the story and was able to see the pictures. They therefore considered it likely that this situation heightened the tendency of young children to engage in a good deal of pointing Ð both in gesture and in the use of deictic expressions (Berman and Slobin, 1994, pp. 24). When they devised a new procedure aimed at overcoming the problem of shared knowledge and perception, they found a considerable reduction in deixis, both to protagonists (‘this one) and to pictured locations and situation (‘here) (Berman and Slobin, 1994, pp. 25). The experimental set-up may therefore be assumed to make the tendency for children to rely on deixis less likely. The children’s narratives were recorded with video and audio and then transcribed and coded for relevant features. In the interest of space we omit details of the transcription process, but these details are available from the authors upon request. 4.3 Results In the justification of their methodological set-up, Berman and Slobin admitted that quantitative analyses proved inadequate in the context of the functionally oriented thrust of their study: […] when we started to evaluate use of tense and aspect in our transcripts, it became evident that the quantitative distribution of these forms needed to be evaluated in terms of how they function within a given text, in relation to

123

480

Michiel van Lambalgen et al.

such factors as overall-tense selected for a particular narrative, and where tense and/or aspectual markings were changed across that text (Berman and Slobin, 1994, pp. 24). In other words, ideally each form is to be analysed within the discourse context. The downside to such a qualitative method is that the numbers (of both subjects and forms) often become too small to perform statistical tests. When groups are further divided according to dominant tense used or using a particular form at all, the numbers become even smaller. For instance, above (in Sect. 3.3.2) we have explained why it makes sense to look at the present perfect (e.g. ‘have looked’) and past perfect (e.g. ‘had looked’). However, the perfect forms make up only 5.1% of the total number of tenses used by both the ADHD and the control group. When we further divide the perfect according to the function the form fulfils within discourse, the numbers become smaller and smaller. Nevertheless, the main hypothesis, (9) above, is corroborated: in their narratives, children with ADHD use a complexity-reducing strategy significantly more often than controls ( p < 0.000). We will give the data in Sect. 4.3.6, but in order to understand the figures given there, it is useful to first study the specific strategies in detail. 4.3.1 Direct Speech Children with ADHD do indeed tend to use more quotes in their narratives. Here are two ways of narrating the same scene: the boy climbs into a tree and looks into a hole, whereupon an owl appears in the hole. (10) a. En die jongen ging zoeken in de boom. En toen zag die een uil. En toen valt ’ie van de boom. And that boy started looking in the tree. And then he saw an owl. And then he falls from the tree. [8yrs, CG]16 b. “Oh nee, ik val!” “Hellup!” “Ga weg, stomme uil, ga weg!” “Oh no, I’m falling!” “Help!” “Go away, stupid owl, go away!” [9yrs, ADHD] We have seen in Sect. 2 that there are definite computational advantages (for the speaker) to using direct speech, and we find indeed that children with ADHD tend to economise in this way (see Table 2). As we can see from Tables 2 and 4 out of 26 children from the ADHD group use direct speech in more than 12% of their utterances, where 12% is the mean use of utterances in direct speech per child, plus one standard deviation. This difference is statistically near-significant ( p = 0.065).17 In Sect. 4.3.6 we shall see that it is advantageous to view such above-average use as a strategy. By combining several such strategies we will be able to construct a highly significant difference between the ADHD and control groups. 16 The child makes a mistake in combining present tense ‘valt’, which could be interpreted as a narrative present heightening the tension, with the adverbial ‘en toen’, which needs a past tense. 17 In the presentation of results, all significance levels are relative to the (non-parametric) Mann-Whitney

test.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD) Table 2 Use of direct speech

481

Percentage of utterances in DS (%)

Control group (%)

ADHD group (%)

0–12 >12 Total

31 (91.2) 3 (8.8) 34 (100)

19 (73.1) 7 (26.9) 26 (100)

4.3.2 Tenseless Utterances As we observed in Sect. 2 the most drastic way to simplify the goals corresponding to tense is to leave out these goals altogether. Much to our surprise, a considerable number of the ADHD children indeed resort to this strategy, also outside direct speech. For example, only 2.9% of the control children used tenseless utterances (outside direct speech) at least 3 times, whereas 19.2% of the ADHD children did so. Overall, the difference between control children and children with ADHD has p-value 0.081, which is striking in view of the small number (66) of tenseless utterances. Here is an example of this phenomenon: (11) En hij is vroeg op. En wat ziet die daar? Kikker verdwenen. And he is up early. And what does he see there? Frog gone. [7yrs, ADHD] The last clause contains a past participle but no tensed verb, and therefore does not require processing of a goal. 4.3.3 Tense Shifts Leading to Incoherence Let us start with an example of discourse incoherence due to a tense shift: (12) Toen waren ze wakker. Toen zagen ze geen kikker meer. Kijkt ie in zijn laars. Toen roepte die. Then they were awake. Then they didnot see a frog anymore. He looks in his boot. Then he called. [7yrs, ADHD] This is incoherent because the event ‘he-look-in-his-boot’ is anchored to utterance time, not to the events described in the preceding and following sentences. Thus the hearer must engage in a recomputation, perhaps re-interpreting ‘He looks in his boot.’ as ‘On this picture, he looks in his boot.’ We expected the children with ADHD to employ such erratic tense shifts much more often. Indeed, we do find such a trend, but the difference does not reach significance. The results are given in Table 3, which shows the number and the percentage of children who commit 0, 1–3, or more than 3 erroneous tense shifts. 4.3.4 Anchoring and Connectivity Berman and Slobin have studied in great detail the development of the linguistic devices used to link events to specific times or to other events. They write (Berman and Slobin, 1994, pp. 173): “Here we are concerned with the way speakers link parts of their narrations to construct a unified, organised piece of text.” They

123

482 Table 3 Tense shifts leading to incoherence

Michiel van Lambalgen et al. Number of shifts

Control group (%)

ADHD group (%)

0 1–3 >3 Total

15 (45.5) 15 (45.5) 3 (9.1) 33 (100)

6 (27.3) 13 (59.1) 3 (13.6) 22 (100)

found (Berman and Slobin, 1994, pp. 186) that “the 5- and 9-year-old English narrators […] rely heavily on ‘and then’ as a means for chaining one clause after another in narrative sequence. Only adults use a variety of linking devices […] .” The narrations of their 5- and 9-year-olds are still strongly additive in character (i.e., using ‘and then’), whereas the adults “stack connectivity devices across chunks of clauses, creating longer stretches of event packaging than noted among children (ibidem, pp. 180)”. Examples include anchoring adverbials like ‘in the morning’. Such examples are used to show that "mature narrators can organise their texts from a thematicallymotivated perspective (ibid.)”. Berman and Slobin finally remark that “[t]hese forms are also found in the childrens narratives, but only occasionally across subjects and texts (ibid.)”. Thus the age group at our disposal is not optimal for studying the phenomenon of interest. Indeed we do not see much difference between the groups in the following domains: – – – –

explicit setting of the reference time (‘in the night’) packaging several events in one utterance (e.g., using ‘when’) use of thematic connectivity markers (‘because’, ‘therefore’, ‘but’…) use of temporal adverbials not fixing reference time (‘again’, ‘a long time’).

In fact, in both groups the vast majority of connectivity markers are of the additive kind (‘and’, ‘and then’, ‘and now’), just as in Berman and Slobin’s data for this age category, and this swamps the greater use of thematic connectivity markers observed in the control group. There is one somewhat curious significant difference in the use of temporal adverbials such as (the Dutch equivalents of) ‘suddenly’, ‘all of a sudden’, ‘immediately’, ‘at once’ …which refer to an abrupt transition between two events, the second of which also occurs unexpectedly. In example 4.3.4 the first event is that of the frog being in the jar, which ends abruptly as the frog escapes. (13) Het was nacht. En Piet sliep. Zijn hond sliep op zijn bed. Opeens stapte de kikker uit de fles. It was night-time. And Piet was asleep. His dog slept on his bed. Suddenly the frog stepped out of the bottle. [8yrs, CG] Such adverbials help the hearer in setting up the event structure, for instance because it introduces a distinction between background and foreground events; one can see this clearly in example (13). In view of this function it is of some interest that if one defines a corresponding strategy for control children18 as ‘use ‘suddenly’ (or its 18 This strategy increases the computational load of the speaker, and eases that of the hearer.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

483

Fig. 3 Pictures 22–24 from “Frog, where are you?”

near-synonyms) three times or more’19 then 15% of the controls employs this strategy, and none of the children with ADHD; this difference is significant ( p = 0.043). 4.3.5 The Perfects We expected significantly less use of the perfects by children with ADHD, in particular of the past perfect. This is not quite what we observed. Overall, children with ADHD used more present perfects than the control children, but fewer past perfects; for the past perfect the difference is near-significant ( p = 0.062). If we look at individual children and count how often they use a past perfect, we do get a striking difference: 38% of control children use the past perfect more than two times,20 against 11.5% of the children with ADHD, a significant difference ( p = 0.041). One reason for this is that the ADHD children use the past perfect much less often in the second half of the story. Consider for instance the final pictures of the story (reproduced in Fig. 3), where the past perfect would be highly appropriate. The boy and his dog come to a log, and behind this log they see the frog. It turns out the frog has found a mate, with whom he has made many little frogs. The reader will have noticed that we just reproduced the story in the narrative present, but using the past tense one would get something like this; (14) En toen zagen ze twee kikkers. En toen veel meer. En toen hadden ze de kikker teruggevonden. And then they saw two frogs. And then many more. And then they had found the frog again. [7yrs, CG] Now compare this to the opening scenes of the story, given in Fig. 2. Setting the scene requires either the past perfect (15) (if the dominant tense of the story is going to be the simple past), or the present perfect (16) (if the dominant tense is going to be the narrative present): 19 This number is determined as the mean plus one standard deviation. 20 This number is determined as the mean plus one standard deviation.

123

484

Michiel van Lambalgen et al.

Table 4 Time-course of the use of the perfects

Diagnosis

Pres. perf. 1–3 Pres. perf. 22–24 Past perf. 1–3 Past Perf. 22–24

Normal

ADHD

9 5 16 18

13 5 12 3

(15) Stefan en de hond die hadden een kikker gevangen. Die had die in de pot gedaan. En die zat stevig vast. Stefan and the dog had caught a frog. He had put it in a jar. And it was firmly shut. [9yrs, CG] (16) Het is nacht. Jan zit naar zijn kikker te kijken die hij gevangen heeft en die hij in zijn potje heeft gestopt. It is night-time. Jan is looking at his frog that he has caught and that he has put in his jar. [9yrs, ADHD] What is interesting here is that whereas in the opening scenes 1–3 there is no significant difference between ADHD and CG with regard to either present or past perfect, the closing scenes elicit significantly more ( p = 0.016) past perfects from the controls; see Table 4. Semantically there is no reason for this difference between first and last pictures, and one may therefore hazard the hypothesis that the computational burden of the past perfect affects the subject only after a while. 4.3.6 Combining Strategies The starting hypothesis of this investigation was (9), repeated here for convenience: (17) Children with ADHD use a complexity-reducing strategy significantly more often than control children. So far we have looked at specific strategies to reduce the complexity of tense processing. We now take a more global view, and look at the children who apply one or more complexity-reducing strategies. For example, a child may use up all his computational resources by avoiding direct speech, thereby producing, say, more erratic shifts in the perfect. Both in case of excessive use of direct speech and of erratic tense shifts the hearer must work hard to construct a coherent story, even though he may not understand why he has to work so hard. Thus, taking the point of view of the hearer, what is necessary is a general definition of complexity-reducing strategy, incorporating the more specific strategies discussed above. Motivated by the analyses given above, we define the overall complexity-reducing strategy of a child as consisting of three components. A child employing one of these component strategies contributes a point (or sometimes 0.5 point) to the number of points of its group. The cut-off values mentioned (5%, 9%, 12%) are determined as the mean plus one standard deviation. (1) Avoiding perfect tenses We speak of a strategy when the child uses no perfect tense at all (1 point).

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

485

Table 5 Distribution of complexity-reducing strategies

Strategy 1 Strategy 2 Strategy 3 Total

Control group (N = 34)

ADHD group (N = 26)

p-value

2 points 0 points 2 points 3 points 7 points

6 points 3 Points 3.5 points 7 points 19.5 points

0.054 0.003 0.136 0.065 0.000

2 points

6.5 points

0.006

(2) Avoiding tensed verbs We speak of a strategy when the child (a) uses tenseless utterances (no direct speech) in more than 5% of his or her narration (0.5 point) (b) uses tenseless utterances (direct speech) in more than 9% of his or her narration (0.5 point). (3) Preference for direct speech We speak of a strategy when the child uses complete direct speech utterances (tensed or tenseless) in more than 12% of his or her narration (1 point). We then get the pattern of results as given in Table 5. We see that the difference between the groups is now highly significant. Thus, the general hypothesis (17) seems to receive some confirmation, even though some of the more detailed hypotheses, such as those about anchoring and connectivity, were not supported, at least in this age group. 5 Conclusion: What can a Formalism do for You? Let us first summarise the results obtained. We presented our main hypothesis as: ‘children with ADHD use strategies for discourse production which reduce the computational burden on themselves, but increase that of the hearer’. We then identified several areas where we expected to see such strategies at work, mainly involving tenses (here including the perfects) and connections between utterances. In the latter domain we observed hardly any differences between children with ADHD and controls. This is probably due to lack of statistical power; it is necessary to investigate older children to see whether there really is a null-effect. In the former domain we found several differences that are near-significant; combined into a general complexity-reducing strategy as given in Sect. 4.3.6 the difference between the groups turned out to be highly significant. What we should conclude from this for ADHD is not an entirely straightforward matter, since there is so much comorbidity in our population. But this is a feature that affects research on ADHD in general. The predictions concerning the use (or non-use) of verb tenses in ADHD were derived from a formal model of tense production and comprehension involving satisfaction of complex goals, together with neuropsychological evidence indicating difficulties with goal maintenance and/or planning toward that goal. The formal model is responsible for the specificity of the predictions. Without the formal model, but equipped only with, say, Trabasso and Stein’s general characterisation of narrative as governed by a hierarchy of goals (Trabasso and Stein 1994), one expects some

123

486

Michiel van Lambalgen et al.

breakdown in the coherence of the story, as was indeed found by Purvis and Tannock (1997). The formal model allows one to be more specific about the computational cost of the devices used to ensure discourse coherence. The model thus acts as a searchlight that allows one to see phenomena one would not have thought of otherwise. One final note of caution. The predictions on the use of verb tense in ADHD were not derived from DSM-type diagnostic criteria, but rather from an EF deficit that is known to occur in a considerable number of cases. A much more direct test of the theory proposed here would thus be to take a sample of subjects known to have difficulties on the Shue and Douglas Go/NoGo task and have them do the narration task. Willcutt et al. (2005) claim that not all of ADHD can be explained through EF deficits. Given the fluidity of the ADHD diagnosis this shouldn’t be a cause for surprise, but it indicates that one ‘all’ children with ADHD to suffer from the problems reported here. References Aarssen, J. (1996). Relating events in two languages. Acquisition of cohesive devices by Turkish-Dutch bilingual children at school age. Tilburg, Netherlands: Tilburg University Press. Baggio, G. (to appear). A semantic approach to the electrophysiology of tense. Language Learning. Berman, R. A., & Slobin, D. I. (Eds.) (1994). Relating events in narrative. Hillsdale, N.J: Lawrence Erlbaum. Blankenstijn, C., & Scheper, A. (2003). Language development in children with psychiatric impairment. Published version of PhD thesis, LOT, Utrecht. Doets, K. (1994). From logic to logic programming. Cambridge, MA: The MIT Press. Kowalski, R. A. (1979). Logic for problem solving, volume 7 of Artificial Intelligence Series. North-Holland, New York and Oxford. Kowalski, R. A. (1995). Using meta-logic to reconcile reactive with rational agents. In Meta-logics and logic programming. pp. 227–242. MIT Press Mayer, M. (1969). Frog, where are you? Dial books for young readers. New York. Ozonoff, S., Strayer, D. L., McMahon, W. M., & Filloux, F. (1994). Executive function abilities in children with autism and tourette syndrom: an information-processing approach. Journal of Child Psychology and Psychiatry, 35, 1015–1032. Pennington, B. F., & Ozonoff, S. (1996). Executive functions and developmental psychopathology. Journal of Child Psychology and Psychiatry, 37, 51–87. Pijnacker, J., Geurts, B., van Lambalgen, M., Buitelaar, J., Kan, C., & Hagoort, P. (2007). Conditional reasoning in high-functioning adults with autism (Submitted). Purvis, K. L., & Tannock, R. (1997). Language abilities in children with attention deficit disorder, reading disabilities, and normal controls. Journal of Abnormal Child Psychology, 25(2), 133–144. Roelofs, M. (1998). “Hoe bedoel je?” De verwerving van pragmatische vaardigheden. The Hague, Netherlands: Holland Academic Graphics. Russell, J. (2002). Cognitive theories of autism. In J. E. Harrison, & A. M. Owen (Eds.), Cognitive deficits in brain disorders (pp. 295–323). Dunitz, London. Shue, K. L., & Douglas, V. I. (1992). Attention deficit hyperactivity disorder and the frontal lobe syndrome. Brain and cognition, 20(1), 104–124. Steedman, M. (1997). Temporality. In J. van Benthem, & A. ter Meulen (Eds.), Handbook of Logic and Language.(pp. 895–938 Chap. 16). Elsevier Science. Stenning, K., & van Lambalgen, M. (2007). Logic in the study of psychiatric disorders: executive function and rule-following. Topoi, 26(1), 97–114. Special issue on Logic and Cognitive Science. Trabasso, T., & Stein, N. L. (1994). Using goal-plan knowledge to merge the past with the present and future in narrating events on line. In M. H. Haith, J. B. Benson, R. J. Roberts, & B. F. Pennington (Eds.), The development of future-oriented processes, (pp. 323–352). University of Chicago Press. van der Meulen, E. M. (2003). Attention-deficit hypercativity disorder in Dutch children. A family study on genotype, phenotype and environment. Wageningen, Netherlands: Ponsen & Looijen.

123

Discourse Processing in Attention-Deficit Hyperactivity Disorder (ADHD)

487

van Lambalgen, M., & Hamm, F. (2004). The proper treatment of events. Blackwell: Oxford and Boston. Willcutt, E., Doyle, A. E., Nigg, J. T., Faraone, S. V., & Pennington, B. F. (2005). Validity of the executive function theory of attentiondeficit/hyperactivity disorder: A meta-analytic review. Biological Psychiatry, 57, 1336–1346.

123