Specific impairments of rule induction in different frontal lobe subgroups

https://doi.org/10.1016/j.neuropsychologia.2004.06.008Get rights and content

Abstract

The neural correlates of inductive reasoning are still poorly understood. In order to explore them, we administered a revised version of the Brixton test [Cortex 32 (2) (1996a) 241], a rule attainment task, to a group of 40 patients with a focal frontal brain lesion of mixed aetiology and to 43 control subjects. To interpret an impairment on the test as suggesting an inductive reasoning deficit a number of alternative hypotheses need first to be considered, namely whether the Brixton impairment could be explained by: (i) a working memory deficit; (ii) a monitoring deficit; (iii) a difficulty in applying an already induced rule; (iv) greater impulsivity. The patients with left lateral (LL) frontal lesions were significantly impaired on the Brixton test; more importantly they were the only group in which none of the alternative hypotheses we explored proved able to explain the flawed performance. In sharp contrast, right lateral lesion patients did not make significantly more errors on the Brixton test than controls, but they produced three times more capture errors (a sign of impaired monitoring processes). The results were interpreted as suggesting functional dissociations between inductive reasoning, monitoring and working memory and a localisation of key processes for induction in left lateral frontal cortex and in right lateral cortex for monitoring and checking.

Introduction

Reasoning is the activity of generating and evaluating arguments. Theories of reasoning distinguish, on the basis of the relationship that holds between premises and conclusions, two main kinds of inference: induction and deduction. An inference is a deduction if the conclusion must be true whenever all the premises are true. Consider, for example, the premises “All men are mortal” and “Socrates is a man” and the conclusion “Socrates is mortal”. As the latter statement must be true whenever both the former ones are, the inference at hand is a (valid) deduction; by contrast, the premises “Socrates is mortal” and “Socrates is a man” do not entail “All men are mortal”; in the latter case the premises provide only limited grounds for accepting the conclusion: these kinds of inferences are called inductions (Rips, 1999). Induction can also be defined as any process of thought yielding a conclusion which increases the semantic information contained in its premises (Johnson-Laird, 1993).

Recently, there has been a growing interest in elucidating the neuroanatomy of the inductive reasoning processes. A series of imaging studies have been devoted to this issue (Duncan et al., 2000; Goel & Dolan, 2000; Goel, Gold, Kapur, & Houle, 1997; Osherson et al., 1998; Parsons & Osherson, 2001; Strange, Henson, Friston, & Dolan, 2001). In most of these, the frontal lobes, particularly their lateral aspects, are activated while the participants are carrying out inductive inferences; but they are often part of a large network of areas.

Many putative processes may be involved in the carrying out of the cognitive operations necessary when any real-life induction occurs. Information must be comprehended and held in working memory; checking may or may not occur. All of these may involve multiple subprocesses. The determination of what the relevant subprocesses are and where they are localised is a complex task, as in our current state of knowledge processes that are not yet tightly definable or operationalisable may be relevant. In addition, the necessary complexity of inductive reasoning tasks often makes it hard to carry out a task analysis. Since, it is difficult to ensure that the processing of all stages other than any hypothetical critical one remain constant across conditions, factorial designs are, in general, difficult to apply using the imaging methodology (but see Duncan et al., 2000).

To determine whether processes necessarily involve a particular region, neuropsychological studies (as other inactivation techniques such as TMS) can provide an important source of evidence (D’Esposito & Postle, 1999; Goel, in press; Price, Mummery, Moore, Frakowiak, & Friston, 1999).

Moreover, with tasks requiring a considerable number of high-level processes, the results of lesion studies are generally easier to interpret functionally as it is not necessary to characterise in detail processes that are unimpaired. Moreover one has (additional) evidence on function from, for instance, the nature of errors as well as from observed dissociations. This provides a second reason for also using the lesion methodology.

A number of tasks, used in lesion studies, have an inductive component. For instance fluid intelligence tests, such as the Raven test (Basso, Capitani, Luzzatti, & Spinnler, 1981; Gainotti, D’Erme, Villa, & Caltagirone, 1986) or the culture fair intelligence test (Duncan, Burgess, & Emslie, 1995) have a major inductive component (Carpenter, Just, & Shell, 1990). However, the structure of the tests makes it difficult to isolate the inductive component neuropsychologically and the variety of item types does not allow a quantitative error analysis to be simply carried out. Theoretically, it would be necessary, in order to assess the inductive component, to administer, along with a fluid intelligence tests, other tests that can assure one of the integrity of the other components. Duncan et al. (1995) indeed used a related procedure. However, as only three subjects were tested, the number of patients studied was too few to strongly sustain any localisation claim. Concept attainment tasks, such as the Weigl test, the Wisconsin card sorting test (WCST) (Drewe, 1974, Milner, 1964, Stuss et al., 2000) and the Brixton spatial rule attainment test (Burgess & Shallice, 1996a) also have an inductive component. The Wisconsin card sorting test is the best known clinical signature of frontal lobe dysfunction. The “discovery” part of the test may indeed stress the inductive competence of cognitively impaired subjects. However, in order to attain normal performance, other abilities are involved, in particular attentional switching, monitoring and sensitivity to negative feedback. In fact, there is some evidence that suggests that the failure observed in frontal patients may arise from impairments to processes other than the inductive one. Thus, perseverative errors, the more distinctive feature of the difference between the performance of frontal patients and that of controls, suggests—although it does not necessitate—a central role for a switching deficit. Moreover, even if frontal patients are told which the relevant criteria of classification are, they can still have pathological performance (Stuss et al., 2000). This relates to the clinical observation that patients often verbalize the three sorting criteria but are unable to use this knowledge effectively (Stuss et al., 2000).

A recently proposed rule attainment task, the Brixton test (Burgess & Shallice, 1996a), is likely to be better suited as a measure of inductive competence in patients. In this test, the participant is presented with a card containing a 2 × 5 display of circles of which one only is filled. The participant must predict where the circles would be completed on the next card. Nine simple rules are used each of which is in operation between three to eight trials. In the Brixton test, the rules which have to be attained pertain to the relation among succeeding stimuli. Thus, the inductive process will be more stressed than on the WCST where the rules directly relate to perceptual features on single cards. In addition, the stimuli will be less prone to automatically trigger overlearned stimulus–response associations and so are less liable to induce perseverative behaviour (see Burgess & Shallice, 1996a). This means that a possible deficit in induction will be less contaminated by other factors. Finally, the variety of different rules used allows a richer error analysis. In the study of Burgess and Shallice (1996), frontal patients as a group both showed a pathological level of performance on the Brixton test, but did not produce a significantly larger number of perseverative errors. Posterior patients, by contrast, performed at a similar level to controls on both measures. This result suggests that the frontal lobes could have a crucial role in inductive reasoning. However, a number of alternative possibilities need to be considered for the pattern of results shown by the patients with frontal lesions:

  • (i)

    A low score on the Brixton test could be due to a working memory deficit (Baddeley, 1997). To carry out any kind of inference, in fact, it is necessary to be able to hold the relevant information in mind. Moreover, it is generally acknowledged that some components of the system underlying working memory performance rely on frontal lobe networks and particularly their ventrolateral and dorsolateral aspects (D’Esposito & Postle, 1999; Owen, Downes, Sahakian, Polkey, & Robbins, 1990; Paulesu, Frith, & Frackowiak, 1993; Petrides, 2000). Thus, it cannot be excluded that the deficit on the Brixton test observed in frontal patients could be generated by a more basic impairment in holding information online. This possibility could not be ruled out given the conditions used in the original study of Burgess and Shallice (1996).

  • (ii)

    In WCST, Stuss et al. (2000) have attributed loss of set errors and perseverative errors in patients with right frontal lesions to a problem of sustained attention or monitoring. An impairment of monitoring and checking could also affect the performance of the Brixton test. Checking and monitoring processes (Burgess & Shallice, 1996b) may also be necessary to satisfactorily evaluate and verify putative newly generated rules or schemas in Shallice & Burgess's (1996) terminology. Recently, Henson and his collaborators have proposed that this process can be localised in the right dorsolateral prefrontal cortex. Their primary evidence was derived from fMRI studies using episodic memory paradigms (Henson, Shallice, & Dolan, 1999; Henson, Shallice, Josephs, & Dolan, 2002 but also see Fletcher & Henson, 2001; Shallice, 2002 for reviews and Petrides, 2000 for an alternative approach to monitoring). For instance, Fletcher, Shallice, Frith, Frackowiak, and Dolan (1998) found that retrieval of a long list of items where the subject needs to monitor his or her output for repeats (Stuss et al., 1994) activates right dorsolateral prefrontal cortex much more than does the equivalent amount of retrieval of one-off paired associates. Can one, however, obtain direct neuropsychological evidence, as these imaging studies suggest, that monitoring or checking processes are relatively lateralised in the frontal cortex?

  • (iii)

    Patients could fail on the Brixton test because of an inability to apply a rule they had already induced. For instance Strange et al. (2001), in their efMRI study on explicit abstract rule induction, suggested that the left dorsolateral frontal cortex is necessary for rule application per se.

  • (iv)

    Finally, a greater impulsivity leading to excessively rapid responding (Burgess & Shallice, 1996a; Miller, 1985, Miller, 1992; Miller & Milner, 1985) could be responsible, in some patients, for the higher number of errors.

Our aims in the current study were three-fold. First, it was to obtain a more precise localisation of regions within prefrontal cortex which give rise to impaired Brixton performance. Secondly, it was to examine whether any such impairment could be explained as a result of malfunctioning of any of the non-induction processes discussed above used in the carrying out of the task. If they are intact, it makes it more plausible to attribute any impairment on the task to malfunctioning of more basic processes involved in induction. Thirdly, we wished to examine using a paradigm different from the imaging paradigms involving episodic memory whether there was any evidence for differential lateralisation of organisational processes on the one hand from checking ones on the other.

As far as localisation is concerned, a procedure developed by Stuss, Alexander and their co-workes has been applied which allows one to produce a somewhat finer localisation of prefrontal impairment than that of simply compare unilateral left and right frontal patients (e.g. Stuss et al., 1998, Stuss et al., 2000). It does so by creating groups that are reasonably coherent and sizeable given the natural history of lesions affecting the frontal cortex, while at the same time making parts which plausibly relate to functional divisions (i.e. lateral versus medial). Following Stuss and collaborators, four subgroups were used: left (LL) and right (RL) lateral, superior (SM) and inferior (IM) medial. However, because of some localisation claims related to induction (Strange et al., 2001) and monitoring (Carter et al., 2000), in a subsidiary analysis we also checked the possible effects, only on these two functions, of fronto-polar (FP) or anterior cingulated (AC) damage, respectively.

We examined the additional processes discussed above that might be involved in performing the Brixton test by using a variety of procedures. First, to carry out the induction of a rule the subject must actively hold information on a sufficient number of cards in mind, as the information will not be stored automatically in a phonological or visuo-spatial buffer (Mitchell, 1972; Phillips & Christie, 1977). An additional working memory task using similar material was designed; it involved one card more than the maximum number necessary in order to disambiguate the rules used. Secondly, an extended error analysis was used to examine whether perseveration was a particular problem. Third, the collection of response times enabled us to consider impulsivity. Finally, the issue of checking and monitoring was addressed by using a second version of the Brixton where interfering rules are potentiated and, for each correct rule, the subject has to avoid making a capture error, which would occur if they obey the interfering rule rather than the previously acquired one. This created the neuropsychological analogue of a situation occurring in a study involving imaging of episodic memory (Henson et al., 1999) which had produced a right dorsolateral activation. Participants had to decide of an item recognised as familiar whether it occurred at precisely the same spatial position and list as when presented. It therefore allowed for possible differential lateralisation of monitoring and checking processes to occur.

Section snippets

Participants

Forty patients with a single focal brain lesion as determined by a CT or an MRI scan were recruited from the Neurological and Neurosurgical ward of Ospedale Civile in Udine (Italy); all patients gave their consent to participate in the study. The aetiology was mixed: stroke, traumatic brain injury and neoplasm (Table 1). Exclusion criteria were: the presence in the clinical history of psychiatric disorders, substance abuse or previous neurological disease, neuroradiological evidence of diffuse

Demographic factors

Education is the only factor significantly affecting FirstHalf score in controls [R2 = 0.11, F (1, 39) = 6.38, P < 0.05 two-tailed] while in patients age is the only one [R2 = 0.44, F (1, 36) = 33.81, P < 0.001 two-tailed].

Days from onset, dimension of the lesion, aetiology and oedema

We evaluated the presence of an effect of each of these four variables on FirstHalf score. For the first two variables, we performed a regression analysis with age, years of education and sex as covariates and FirstHalf score as a dependent variable. In none of the cases was

Discussion

The Brixton spatial rule attainment task (Burgess & Shallice, 1996a) is a procedure devised to investigate impairments in rule induction and rule following. It is less prone to perseverative types of responding than the WCST, but still produces deficits following prefrontal lesions. A key process in the Brixton test is held to be rule induction. However, a number of other processes/abilities may be required for satisfactory task performance. We considered the following processes, all of which

Acknowledgements

We are grateful to Dr. Serena D’Agostini for her kind help in collecting the patients’ scans.

References (43)

  • A.M. Owen et al.

    Planning and spatial working memory following frontal lobe lesions in man

    Neuropsychologia

    (1990)
  • J. Sandson et al.

    Varieties of perseveration

    Neuropsychologia

    (1984)
  • D.T. Stuss et al.

    Wisconsin card sorting test performance in patients with focal frontal and posterior brain damage: Effects of lesion location and test structure on separable cognitive processes

    Neuropsychologia

    (2000)
  • A. Baddeley

    Human memory: Theory and practice

    (1997)
  • A. Basso et al.

    Intelligence and left hemisphere disease. The role of aphasia, apraxia and size of lesion

    Brain

    (1981)
  • P.W. Burgess et al.

    Confabulation and the control of recollection

    Memory

    (1996)
  • P.A. Carpenter et al.

    What one intelligence test measures: A theoretical account of the processing in the Raven Progressive Matrices Test

    Psychological Review

    (1990)
  • C.S. Carter et al.

    Parsing executive processes: Strategic vs. evaluative functions of the anterior cingulate cortex

    Proceedings of the National Academy of Sciences of the United States of America

    (2000)
  • H.D. Critchley et al.

    Human cingulate cortex and autonomic control: Converging neuroimaging and clinical evidence

    Brain

    (2003)
  • J. Duncan et al.

    A neural basis for general intelligence

    Science

    (2000)
  • P.C. Fletcher et al.

    Frontal lobes and human memory: Insights from functional neuroimaging

    Brain

    (2001)
  • Cited by (0)

    The study was carried out in the Santa Maria della Misericordia Hospital (Udine, Italy) and in SISSA (Trieste, Italy).

    View full text