Problem-solving and spatial working memory in patients with ...

Schizophrenia Research 37 (1999) 21–33

Problem-solving and spatial working memory in patients with schizophrenia and with focal frontal and temporal lobe lesions T.M. Rushe a,*, R.G. Morris b, E.C. Miotto b, J.D. Feigenbaum b, P.W.R. Woodruff a, R.M. Murray a a Department of Psychological Medicine, Institute of Psychiatry and King’s College School of Medicine, De Crespigny Park, London, SE5 8AF, UK b Department of Psychology, Institute of Psychiatry and King’s College School of Medicine, De Crespigny Park, London, SE5 8AF, UK Received 10 December 1997; accepted 25 September 1998

Abstract Problem-solving ability was investigated in 25 DSM-IIIR schizophrenic (SC ) patients using the Tower of Hanoi (TOH ) task. Their performance was compared to that of: (1) 22 patients with neurosurgical unilateral prefrontal lesions, 11 left (LF ) and 10 right hemisphere (RF ); (2) 38 patients with unilateral temporal lobectomies, 19 left (LT ) and 19 right (RT ); and (3) 44 matched control subjects. Like the RT and LF group, the schizophrenics were significantly impaired on the TOH. The deficit shown by the schizophrenic group was equivalent whether or not the problems to be solved included goal–subgoal conflicts, unlike the LF group who were impaired specifically on these problems. The nature of the SC deficit was also distinct from that of the RT group, in that the problem-solving deficit remained after controlling for the effects of spatial memory performance. This study indicates, therefore, that neither focal frontal nor temporal lobe damage sustained in adult life is a sufficient explanation for the problem-solving deficits found in patients with schizophrenia. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Schizophrenia; Problem-solving; Spatial memory; Goal–subgoal conflict

1. Introduction The presence of neuropsychological impairment in schizophrenia was recognized early in the seminal work of Kraeplin (1919). Although the cause for this remains unclear, Kraeplin’s belief that the loss of the ‘critical’ faculty in problem-solving and planning relates to frontal lobe dysfunction has in * Corresponding author. Tel: +44 171 703 5411; Fax: +44 171 701 9044.

recent years received some empirical support. In particular, cognitive activation studies that have employed tasks that are thought to be dependent upon the integrity of the prefrontal cortex [for example, the Wisconsin Card Sorting Test ( WCST ) ( Weinberger et al., 1991, Kawasaki et al., 1993), verbal fluency tasks (Lewis et al., 1992; Curtis et al., 1998) and the Tower of London test ( TOL) (Andreasen et al., 1992)] demonstrated that schizophrenic patients had less task-related frontal activation than healthy volunteers. An explanation of observed executive function

0920-9964/99/$ – see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S0 9 2 0 -9 9 6 4 ( 9 8 ) 0 0 12 9 - 7

22

T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

deficits in schizophrenia as solely a consequence of frontal lobe dysfunction is, however, obscured for a number of reasons, not least of which concerns the recognition that most higher-order processes are mediated by complex neural networks. As such, a disruption at any part of that would produce similar deficits (Luria, 1973). For example, hypofrontality, as measured by regional cerebral blood flow (rCBF ), during performance of the WCST has been shown to be strongly correlated to both right and left hippocampal volume ( Weinberger et al., 1992). Such findings are consistent with the proposition that executive dysfunction in schizophrenia is a result of abnormalities in the neural circuitry, which links the frontal and medial temporal lobes ( Weinberger, 1991). More direct support for this comes from a recent PET activation study by Frith et al. (1995), in that they demonstrated that people with schizophrenia failed to exhibit the normal correlated activity between the frontal and temporal lobes during a verbal fluency task. The synchronous activity of the prefrontal cortex and temporal lobes is likely to be required in various problem-solving tasks that require the simultaneous storage and manipulation of information. One such task is the Tower of Hanoi ( TOH ). In clinical versions, this consists of three or four discs of differing sizes slotted into three rods. The aim is to move the discs from a starting arrangement to a ‘goal’ arrangement in as few moves as possible. In order to be successful, it is necessary to plan the sequence of moves, the subject usually delaying initiation of activity whilst a sequence is computed. In certain instances, it may be necessary to explore several solutions and compare them for efficacy, thus increasing the demands on working memory. Schizophrenic patients have consistently shown impairments on this task (Goldberg et al., 1990; Schmand et al., 1992) and comparable tasks, for example, the TOL (Andreasen et al., 1992; Morris et al., 1995). Morris et al. (1995), using a computerized version of the TOL task, which facilitated the measurement of movement times at different stages of the problem-solving process, were able to demonstrate that the problem-solving deficit in schizophrenics was not simply due to either motor

or cognitive slowing. A more recent study by Pantelis et al. (1997) replicated this result in chronic schizophrenic patients, using a similar TOL test. The current study explores further the planning deficit in schizophrenia using a computerized threedisc version of the TOH1. By incorporating data from a previous study by Morris et al. (1997), the performance of the schizophrenic patients was compared directly with that of patients with neurosurgical lesions in the right or left, frontal or temporal lobes. By doing so, our aim was to see whether the cause of the problem-solving deficit in schizophrenia could be dissociated from that in patients with focal lesions in relevant cortical regions. A series of tests based around this task was designed to dissociate different cognitive components of problem-solving: (1) A ‘motor control’ task was employed in order to differentiate between slowness of ‘thinking’ and perceptual motor speed. In this task, the subject is led through a series of moves by the computer, and the resultant response times can be subtracted from the TOH thinking times to obtain ‘purer’ measures of cognitive speed. In previous studies, using the same method with the TOL (Morris et al., 1995; Pantelis et al., 1997), a reduction in planning time had not been found in schizophrenia, and this was also the case in patients with frontal or temporal lobe lesions using the current task (Morris et al., 1997); (2) A Tower of Hanoi ‘memory span’ task was instituted to investigate the ability of the subjects to hold ‘on-line’ sequences of TOH moves, as would be needed in solving a TOH problem. The computer presents a series of 1 The Tower of Hanoi was selected in this instance (opposed to, for example, the Tower of London) because it has a larger state space, thus facilitating the experimental design. For example, in testing spatial memory span, the state space enabled the distance between the starting and end state to be seven moves, rather than six, the maximum on Tower of London. Also, the larger state space increased the potential problems available making possible the systematic comparison of congruent and conflict problems in the current study, and also in ‘follow on’ studies where other factors have been explored (e.g. Morris et al., 1998).

23

T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

moves, which then have to be replicated by the subjects. A deficit in spatial working memory has been suggested in schizophrenia, but would this provide an explanation of a planning deficit in schizophrenia? The study by Morris et al. (1997) had shown spatial memory impairment in left frontal lesion and right temporal lesion patients, but only in the latter was this related to a planning deficit. The involvement of spatial working memory in planning was demonstrated by this study; in order to plan a sequence of moves, it is necessary to hold the plan in memory prior to execution of the moves. Additionally, it may be necessary to construct more than one plan and compare the efficiency of each, in order to solve a particular problem in the minimum number of moves (Morris et al., 1998). Schizophrenic patients show spatial working memory deficits, for example, being impaired on a spatial delayed-response task (Park and Holzman, 1992) and a computerized spatial working memory task used by Pantelis et al. (1997); (3) The types of TOH problems were split according to whether they contained a ‘goal–subgoal conflict’ (Goel and Grafman, 1995). This relates to the fact that the solution to a TOH problem can be obtained by splitting it up into a number of subgoals. In certain instances, the appropriate subgoal is in a direction away from the final goal, the scenario described as the goal–subgoal conflict ( Kahney, 1986; Garnham and Oakhill, 1994). By comparing problems that maximize and minimize this conflict, it was possible to determine whether this factor was a significant cause of the problem-solving impairment. This was found to be the case with patients who have left frontal lesions, but would it provide an explanation of a putative deficit in schizophrenic patients? The difficulty found with goal–subgoal conflict problems was related by Morris et al. (1997) to a more fundamental impairment in responding to ‘novelty’ scenarios where a pre-existing algorithm or schemata does not exist (Shallice and Burgess, 1996). The prefrontal cortex has also been implicated

Goldberg et al. (1994a,b) in processing cognitively novel situations (Goldberg, 1995). In summary, the study investigated whether people with schizophrenia show a different pattern of impairment that might in turn implicate a different causation of executive impairment, other than a simple lesion model relating to the frontal and temporal cortex.

2. Method 2.1. Subjects 2.1.1. Schizophrenics The study included 25 (20 male, five female) DSM-III-R schizophrenic patients (SC ). Exclusion criteria included a history of hypertension, unconsciousness, head injury, alcohol or intravenous drug abuse. The demographic data of the schizophrenic group, as well as the frontal lobe, temporal lobe and normal control groups are presented in Table 1. The mean age, years spent in education and estimated general intelligence, as assessed by the National Adult Reading Test—Revised (NART—R; Nelson, 1992), did not differ significantly between the groups. The mean age of onset of schizophrenia was 24.5 years (s.d.=6.8; range=12–37), and the mean daily dose of neuroleptic medication for the group, in terms of chlorpromazine equivalents, was 653 (s.d.=694; range=0–2210). The symptom severity was assessed using the Scale for the Assessment of Table 1 Comparison of the six groups by sex, age and IQ, as estimated by the National Adult Reading Test—Revised (mean scores are given with standard deviations in parentheses) Group

Age (years)

IQ (NART )

Education (years)

SC RF LF RT LT CTL

35.5 37.6 31.0 33.9 33.1 36.2

105.9 103.3 107.1 103.7 104.2 107.2

13.24 (3.1) 14.0 (2.5) 12.77 (2.3) Not recorded Not recorded 13.6 (3.6)

(6.84) (5.4) (3.2) (1.9) (2.2) (1.7)

(10.0) (5.1) (3.2) (2.9) (2.1) (1.3)

SC, schizophrenia; RF, right frontal; LF, left frontal; RT, right temporal; LT, left temporal; CTL, controls.

24

T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

Negative Symptoms (SANS; Andreasen, 1982, 1983), and the Scale for Assessment of Positive Symptoms (SAPS; Andreasen and Olson, 1982). The mean score of the patients on the SANS was 5.83 (s.d.=4.59; range=0–15) and on the SAPS 5.47 (s.d.=3.47; range=0–13). Neither positive nor negative symptoms were significantly correlated with either age or duration of illness. In addition, the schizophrenic patients were assessed using a short form of the Wechsler Adult Intelligence Scale—Revised ( WAIS-R) ( Wechsler, 1981; Canavan et al., 1986). Their mean intelligence quotient (IQ) was 97.81 (s.d.=17.06). The neurosurgical patient groups are described fully in Morris et al. (1997) and therefore are described only briefly here. All patients were studied after a minimum of 6 months post-operatively. They included: Frontal Lobe Lesion Group: Twenty-one patients (nine male, 12 female) who had undergone unilateral frontal lobe neurosurgery [11 left (LF ); 10 right (RF )]. The verification of the locus and extent of the lesion were established both by neurosurgeons’ drawings at the time of the operation and by use of CT and MRI scanning. Out of the 11 LF patients, there were seven cases with frontal lobectomies for the relief of intractible epilepsy, two where an astrocytoma had been removed, one case of a discrete lesion due to insertion of a shunt and one case of removal of an arteriovenous malformation. For the RF patients, six had undergone neurosurgical treatment for epilepsy, three underwent removal of a meningioma and one case underwent removal of a glioma [see Morris et al. (1997) for further details]. The Temporal Lobe Lesion Group: Of 38 patients (13 female, 25 male), 19 had undergone unilateral left temporal lobectomies (LT ) and 19 right temporal lobectomies (RT ) for the relief of intractable epilepsy. Standard ‘en bloc’ resection was used (Polkey, 1989), which involved removal of 5.5–6.5 cm of medial temporal tissue from the anterior pole, including the amygdala and approximately the anterior two-thirds of the hippocampus [the seizure characteristics of this sample are described by Feigenbaum et al. (1996)]. Healthy Controls: The performance of the five

patient groups was compared with 44 healthy normal controls (25 female, 19 male) recruited via advertisements in two local job centres.

2.2. Tower of Hanoi Task 2.2.1. Procedure Each task was administered using a Taxan 775 14 monitor fitted with a touch-sensitive screen ( TSS). A stylized representation of the layout is given in Fig. 1. The computerized display consists of two arrangements, using three-dimensional graphics, each of which has three discs of different sizes, ‘slotted’ into ‘rods’, the latter based on a ‘plinth.’ Each rod can hold a maximum of three discs each. The subject is instructed to rearrange the bottom set of discs to make them look like the top set. The discs can be moved by touching them and then the position on the rod that is the preferred destination. In order to signify appropriate selection, a white cross appears on a disc, and the computer emits a tone. A potential move can be cancelled by touching the ‘change mind’ box. Before starting, the three main rules of the task are explained. These include: (1) not being able to

Fig. 1. Examples of the problems used in the experiment. ‘CONGRUENT’ and ‘CONFLICT’ examples are shown for difficulty levels 4 and 5, as used in the experiment. The aim is to shift the bottom ‘discs’ in each array so as to match the top arrangement in the array (a stylized representation is given for illustration purposes: S, small discs; M, medium-sized discs; LL, large discs).

T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

move a disc if a smaller disc is in place above it; (2) not putting a disc into a position when there is nothing directly underneath it, other than the ‘plinth’; and (3) not placing a larger disc on top of a smaller disc. If the subject breaks these rules, the computer simply does not respond. If the subject persists, the experimenter prompts the subject, indicating the rule they are breaking. The subject is instructed at the beginning that he/she should solve the problems taking as few moves as possible. Although the subject can change his/her mind about a move and place the disc back in a previous position, he/she is effectively penalized for doing so, because this increases the number of moves taken overall. The imperative to solve the problem in the minimum number of moves effectively reinforces the task as a test of planning ability. 2.3. Computerized Tower of Hanoi Motor Control The 3-D Computerized Tower of Hanoi Control (3-D CTOH-Control ) provides a baseline measure of perceptual–motor speed. The subject is required to copy a sequence of single moves executed by the computer. On the top half of the screen, the computer changes the arrangement by one disc at a time, and the subject follows suit by moving the equivalent disc in the lower half. At the start of each trial, the two arrangements differ from each other by just one disc. Using this method, the subjects are ‘led’ through the optimum move sequence for all the problems presented in the 3-D CTOH-Test. The subjects are instructed to keep matching the bottom and top arrangements as rapidly as they can, whilst maintaining accuracy of movement. 2.4. Tower of Hanoi Memory Span For the memory task, the 3-D Tower of Hanoi Memory Span (3-D TOH-MS) task, only one arrangement is presented, positioned at the bottom of the screen. The subject views this arrangement, which then moves through a sequence of moves as if solving a particular problem. The arrangement then disappears from view to be replaced with the starting arrangement. The subject then has to

25

move the discs in the same sequence. As in the problem-solving task, he/she has to touch the relevant disc and then the location to which it should move. The computer starts with arrangements that move by only two moves, increasing until it reaches a maximum of seven. There are four trials at each level, and the subject has to respond with at least two correct sequences in order to move on to the next level. 2.4.1. Design For practice on the 3-D CTOH-Test, there were four problems that could be solved in two moves, the first two used to demonstrate the task and the use of the TSS. A further three problems taking three moves were presented for additional practice to ensure mastery of the procedure before going on to the test problems. These were eight problems, half requiring four moves and the other half five moves. In addition, the problems are split equally according to whether they have a goal–subgoal conflict or otherwise (defined according to whether the first move is significantly away from the direction of the final goal ). The two types of problems are labelled ‘CONFLICT’ and ‘CONGRUENT’, respectively, and are interleaved in terms of type, with CONGRUENT problems, followed by CONFLICT, and so on. The distinction is illustrated in Fig. 1. First, in the four-move problem (CONGRUENT ), the smallest disc is moved to the middle peg, in the direction of the destination position. In the second (CONFLICT ), the small disc is moved away from the middle peg, the final site of this disc. In the five-move CONGRUENT problem, the middle-sized disc is moved to the centre, in the direction of the right peg, the final position. In the CONFLICT condition, the middle-sized disc has to be moved initially to the far left, away from the final position on the right. 3. Results 3.1. Accuracy 3.1.1. Number of moves The data from the 3-D CTOH-Test were expressed in terms of the number of moves above

Fig. 2. 3-D Computerized Tower of Hanoi Task showing the mean number of moves above the minimum for the four and five move problems for each of the six groups. Bars represent the standard error of the mean.

26 T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

27

T.M. Rushe et al. / Schizophrenia Research 37 (1999) 21–33

the minimum needed to solve each problem. The data were then split according to the CONGRUENT versus CONFLICT distinction, comparing performance for the four- and fivemove problems (see Fig. 2). A three-way MANOVA with Group as a between-subject factor and Goal/Subgoal Conflict (GSC ), and Problem length (PL) as within-subject factors revealed a main effect of group [F(5,127)=2.90, p