Non-spatial attention disorders in patients with frontal or posterior brain damage

Transcription

1 Brain (1996), 119, Non-spatial attention disorders in patients with frontal or posterior brain damage Olivier Godefroy, Chantal Lhullier and Marc Rousseaux University of Lille Department of Neurological Rehabilitation, France Correspondence to: Dr O. Godefroy, Service de Riiducation Neurologique, F Lille, France Summary The few studies that have looked at attention in patients with brain damage suggest a prominent role for the frontal lobe in nonspatial attentional control. However, the studies usually focus on one variety of attention and do not address the nature of the alteration of attention. In addition, the behavioural consequences of nonspatial attentional deficit remain unknown. The aim of this study was to evaluate the role of focal brain damage on divided and focused attention and the relationship between attention disorders and behavioural changes. The study group consisted of patients with lesions of the prefrontal and posterior cortices and control subjects. The assessment of attention used reaction time tests that evaluated the ability to divide attention between two sources (detection tests) and to focus attention on one source (Go/ No-Go Tests). The response retardation of the 'frontal' group became more pronounced as the number of sources to be monitored increased, suggesting the presence of a deficit of divided attention. Focused attention deficit was demonstrated in the 'frontal' group by the more frequent responses to irrelevant stimuli on Go/No-Go Tests. Both focused and divided attention deficits were prominent when the lesion included the left dorsolateral prefrontal cortex and the caudate nucleus. Selective deficit of divided or focused attention was shown in a few patients. Finally, the clinically assessed distractibility was related to disorders of divided and focused attention. This study provides additional evidence for the prominent role of the frontal lobe in nonspatial attention regulation and shows that it also operates in elementary perceptuomotor processes. The relationship between distractibility and attention indexes supports the idea that attention disorders may have a functional counterpart that is clinically assessable. The demonstration of selective deficit of divided or focused attention suggests that nonspatial attention depends upon different mechanisms and that it is not an undifferentiated, general purpose mechanism. Further studies addressing the nature of the interactions between attentional mechanisms and other cognitive processes are required. Keywords: neuropsychology; attention; control processes; frontal lobe; reaction time Abbreviations: ANOVA = analysis of variance; ANCOVA = analysis of covariance Introduction The presence of attention disorders is frequently claimed in patients with brain damage, especially when the lesion includes the frontal lobe (H caen, 1964; Luria, 1978). Attention deficits are usually demonstrated on clinical grounds or from psychometric evaluation, but formal evaluation of attention capacity is rarely performed in patients (Wilkins et al., 1987). Spatial orienting of attention and hemineglect have been the most frequently addressed disorders in recent years. Conversely, nonspatial attentionaj control has been poorly evaluated. Despite conflicting results, the few studies that have been performed suggest a prominent role of the frontal lobe in attention regulation. A deficit of focused attention (Perret, 1974) and interference control (Drewe, 1975; Salmaso and Denes, 1982; Leimkuhler and Mesulam, Oxford University Press ) has been demonstrated in some studies, but not in others (Stuss et al., 1981; Eslinger and Damasio, 1985; Phillips et al., 1987; Stuss, 1991). The study from Wilkins et al. (1987) showed a deficit of sustained attention, a result that we failed to replicate (Godefroy et al., 1994a). Performance on tasks using cueing demonstrated poor benefit from advance information, suggesting a deficit of selective attention (Alivisatos, 1992). Finally, neurophysiological studies have shown disturbances of electrophysiological correlates of sensory gating and of selective attention in patients with unilateral dorsolateral frontal damage (Knight, 1991). Although these results highlight the role of the frontal lobe in attention regulation, they provide little information on the nature of the attentional modulation supported by the

2 192 O. Godefroy et al. prefrontal cortex. A large variety of impairments are attributed to attention disorders, such as spatial hemineglect, distractibility, aspontaneity, inability to perform two tasks simultaneously, suppression of prepotent response tendencies, planning and problem solving deficits, interference effects, impairment of shifting and engagement of selective sets and vigilance decrements (Hecaen, 1964; Shiffrin and Schneider, 1977; Schneider et al., 1984; Sperling, 1984; Stuss and Benson, 1986; Posner and Petersen, 1990; Stuss, 1991; Airport, 1993; Mangun et al., 1993). As most previous studies have only evaluated one type of attention using a single test, it remains largely unknown whether the reported attention deficits are test-dependent or are an emerging manifestation of a general attentional dysfunction (Stuss and Benson, 1986; Allport, 1992). This question is of practical and theoretical interest. Attention is usually evaluated using complex tests derived from experimental psychology, and the link between the reported deficits and the clinical manifestations usually attributed to attention disorders, such as distractibility, has not been addressed. Such evaluation may determine whether there is a behavioural aspect of nonspatial attention disorders. From a theoretical point of view, the nature of attention phenomena remains unknown, despite its critical implication for the cognitive architecture of the brain. Evidence for the limitation of the processing capacity of the brain has led to the assumption that the source of limitation is attention itself and, therefore, that it is a unitary system (Broadbent, 1958; LaBerge, 1973; Baddeley, 1986; Posner and Petersen, 1990; Allport, 1993). Furthermore, the close link between attention and the processes involved in non-routine operations (Shiffrin and Schneider, 1977; Schneider et al., 1984) has resulted in a conflicting situation that assimilates attentional selection mechanisms to a central regulation system (Norman and Shallice, 1986; Baddeley, 1986; Posner and Petersen, 1990). In contrast with this view, studies examining the nature of nonspatial attentional mechanisms remain rare as noted by Shallice (1988) and Allport (1993). A unitary concept of attention would define attention as an undifferentiated general system influencing numerous cognitive processes. On the other hand, a restrictive concept of attention would predict that attention phenomena should reflect some modulations specific to the task considered (Allport, 1993). One way to address this unresolved theoretical issue is to examine whether brain damage may induce selective deficit of one component of attention (Shallice, 1988). Evidence for selective deficit of a component of attention would disprove the hypothesis of a general causal mechanism. These considerations suggest that, in order to understand further attention disorders, it is first necessary to evaluate the link between the different components of attention and the relationship with behavioural changes. In order to evaluate nonspatial attention capacity in patients with frontal damage, we have conducted studies in patients suffering from ruptured aneurysm of the anterior communicating artery (Godefroy et al., 1994a, b; Godefroy and Rousseaux, 1996). These studies suggest that these patients suffer from deficits of divided and focused attention that are not related to increased fatigability, poor motivation, insufficient training and vigilance decrement. These preliminary evaluations did not compare the performance of the 'frontal' group with that of the patients with damage to the posterior cortex and this prevents conclusions being drawn about the specific role of the prefrontal cortex in attentional modulation. In addition, the link between divided and focused attention has not been addressed. The aim of the present study was to evaluate (i) the effect of focal brain damage on nonspatial attention capacity, (ii) the link between focused and divided attention, and (iii) the relationship between attention disorders and behavioural changes. Subjects The study group consisted of 11 patients with prefrontal damage ('frontal' group) (Godefroy et al., 1994a), 11 patients with vascular damage of the retro-rolandic cortex ('posterior' group) and 20 control subjects free from any CNS disorder. Exclusion criteria for all subjects were previous neurological or psychiatric disorders, sensory or motor disorders, current psychoactive or antiepileptic medication, severe cardiac, respiratory or renal failure, diabetes mellitus, alcoholism or illiteracy. All subjects were right-handed and native French speakers. The three groups did not significantly differ in age or education (Table 1). MRI was used to determine the location and volume of cerebral lesions, according to a method providing a significant inter-rater agreement (lesion location: Kappa statistic K = 0.504, P< ; lesion volume: Kendall coefficient of concordance W = 0.961, P< 0.001). Reconstructions were performed using Talairach and Tournoux's method (1988). Three independent examiners rated the signal intensity using axial slices (T 2 -weighted spin-echo sequences) as 0 (no lesion) or 1 (lesion). The lesion volume was evaluated using the following method: lesion outline was defined on each slice with a mouse-driven cursor. An original software program automatically interpolated the outlines and computed the volume. Results of the neuroradiological study are given in Table 1. The neuropsychological and behavioural assessment was performed earlier (Mann-Whitney U test: P < 0.03) on patients in the 'posterior' group (mean delay: 23±17 days) than those in the 'frontal' group (55±40 days). The general neuropsychological assessment included the Wechsler Adult Intelligence Scale (Wechsler, 1970), the Mini-Mental State Evaluation (Folstein et al., 1975) and the forward digit span. The behavioural assessment evaluated the presence of distractibility and aspontaneity that are usually attributed to attention disorders (H6caen, 1964; Stuss and Benson, 1986; Shallice, 1988; Fuster, 1989). The distractibility scale (1: normal; 2: light distractibility; 3: mild distractibility easily reversed by instructions; 4: distractibility transiently reversed; 5: major distractibility not or poorly reversed despite repeated

3 Attention disorders in focal brain damage 193 Table 1 Details of subject groups Mean age (years) Education level Lesion location Frontal lobe Basal Lateral Caudate nucleus Parietal lobe Temporal lobe Thalamus Lesion size (L/R) (L/R) (L/R) (L/R) (L/R) (L/R) (L) (R) Frontal group (n = 1) Posterior group (n = 11) Control group (n = 20) 55 (±15) 1.8 (±0.7) 6/8 5/6 4/ (±2.89) 7.28 (±11.94) 54 (±17) 1.7 (±0.7) 4/2 9/3 3/1 9.8 (±9.65) 4.2 (±10.8) 50 (±16) 2 (±0.7) Standard deviations are given in brackets. Education level: 1, <9 years of school; 2, 9-11 years; 3, 3=12 years. Lesion location is given as region name (side), number of patient/volume of lesion (cm 3 ); L = left side; R = right side. P = values of the Kruskal-Wallis test. Table 2 Mean scores on behavioural assessment, Mini-Mental State Evaluation (MMSE), Wechsler Adult Intelligence Scale (IQ) and digit span Frontal group Posterior group Control group Aspontaneity Distractibility MMSE IQ Digit span 1.36 (±0.5)* f 2.45 (±1.7)*-t 27 (±4)* 98 (±16) 6 (±2) 1 (±0)t 1.09 (±0.3) t 26 (±2.7)* 92 (±17) 5.6 (±1.7) 1 (±0) 1 (±0) 29 (±0.9) NP 5.5 (±0.69) Standard deviations are given in brackets. NP = not performed. P = values of the Kruskal-Wallis test: ^significant difference with controls; ^significant difference between patient groups. instructions) and the aspontaneity scale (1: normal; 2: aspontaneity reversed by light stimulation; 3: aspontaneity reversed by stimulations but with latency; 4: aspontaneity reversed with difficulty and transiently; 5: major aspontaneity or akinetic mutism) were used (Godefroy et al., 1994a). The 'frontal' group differed from both other groups in higher behavioural changes, and both patient groups differed from controls in the Mini-Mental State Evaluation (Table 2). UNIMOOAL AUDITORY TEST A A A A UNIMODAL VISUAL TEST v v v IS) v 61 Methods Reaction time tests Attention was assessed using four reaction time tests that evaluated the ability (i) to regulate attention across sensory modalities, and (ii) to focus attention on one modality. All tests used visual and/or auditory successive stimuli at pseudorandomly varying intervals (1, 4, 7 or 10 s). Stimuli were generated by an Apple lie computer equipped with a real-time clock; auditory stimuli were tone sounds (1000 Hz) and visual stimuli were white squares (Leclercq et al., 1988). Thus, there were two possible stimulus combinations that defined two types of test: (i) the unimodal tests that consisted of a block of 32 stimuli of fixed modality within each test (visual stimuli in the unimodal visual test and auditory stimuli in the unimodal auditory test); (ii) the bimodal test that was characterized by a combination of auditory (n = 32) and visual (n = 32) stimuli delivered in a pseudorandom fashion BIMODAL TEST V V A A 61 SI 61 T T T IPSIMOOAL CROSSMODAL IPSIMOOAL Fig. 1 Sequences of stimuli in unimodal auditory and visual tests and in the bimodal test. The unimodal (unimodal tests), ipsimodal (V-V or A-A sequences) and crossmodal (V-A or A-V sequences) situations are illustrated. A = auditory stimulus; V = visual stimulus; ISI = interstimuli interval (randomly varying from 1 to 10 s). (Fig. 1). The block of 64 stimuli in the bimodal test was used for the Go/No-Go Tests. Reaction times were measured under three different conditions: (i) simple unimodal reaction time tests, where

4 194 O. Godefroy et al. Table 3 Mean reaction time [(visual reaction time+auditory reaction time)/2] in the Detection (unimodal and bimodal tests) and Go/No-Go tasks according to the situation (unimodal, ipsimodal and crossmodal), mean commission and omission rates (%) and nonparametric indexes of sensitivity [P(I)] and criterion (B 1 ) Detection Reaction time Unimodal Ipsimodal Crossmodal Go/No-Go Reaction time Ipsimodal Crossmodal Commission (%) Ipsimodal Crossmodal Omission (%) Ipsimodal Crossmodal R' D Ipsimodal Crossmodal Ipsimodal Crossmodal Frontal group 313 (±121) 420 (±268) 482 (±380) 477 (±198) 532 (±168) 16 (±31) 27 (±36) 0.38 (±4) 0.28 (±3) (±0.105) (±0.092) (±0.14) (±0.13) Standard deviations are given in brackets. Posterior group 305 (±60) 346 (±92) 363 (±104) 430 (±111) 446 (±126) 4 (±9) 10 (±21) 3.76 (±12) 1.19 (±8) (±0.058) (±0.045) (±0.070) (±0.053) Control group 268 (±60) 298 (±99) 316 (±105) 394 (±101) 404 (±107) 2 (±6) 13 (±25) 2.05 (±7) 0.61 (±4) (±0.025) (±0.058) (±0.027) (±0.069) subjects responded to all stimuli that were either visual (unimodal visual test) or auditory (unimodal auditory test); (ii) simple bimodal reaction time test, where subjects responded to all stimuli, regardless of its modality (visual or auditory); (iii) Go/No-Go reaction time tests, where subjects only responded to stimuli of a specific modality and ignored stimuli of the other modality (Go/No-Goi: go signals = visual stimuli; G0/N0-G02: go signals = auditory stimuli). The mode of the subject's responses did not differ across conditions: for all tests, reaction times were measured by depression of the same response button by the right index finger and subjects were told to respond as quickly, and as accurately, as possible. The nature of the stimulus presentation differed between the unimodal and bimodal tests: the stimulus modality was fixed in each unimodal test, whereas it changed in the bimodal test. The variation of stimulus modality in the bimodal test generated two different situations: (i) the ipsimodal situation when the preceding stimulus was from the same modality (auditory-auditory or visual-visual sequences); (ii) the crossmodal situation when it was from the other modality (visual-auditory or auditory-visual sequences) (Fig. 1). Both the unimodal and ipsimodal trials were characterized by successive stimuli of the same modality and thus, were indistinguishable. However, they differed with respect to the predictability of the stimulus modality since, in the ipsimodal trials, subjects did not know the modality in advance. Following the approach of Boulter (1977), the slowing of response in ipsimodal trials compared with the unimodal trials may reflect the effect of the allocation of attention to two sensory modalities. We will refer to the response retardation in the ipsimodal situation as an index of divided attention (ipsimodal reaction time-unimodal reaction time). Further changes in reaction time in the crossmodal situation may be related to the shift of attention from one channel to the other (LaBerge, 1973; Rist and Thurm, 1984). The nature of the stimulus presentation did not differ between the bimodal and Go/No-Go Tests, as both tests used the same block of 64 stimuli, but the task was modified simply by changing the instructions given to the subject. Subjects had to direct their attention to the attended modality and suppress any response to stimuli on the unattended modality. Following the approach of Shiffrin and Grantham (1974), the ability to focus attention on one channel was evaluated by erroneous responses to stimuli of the irrelevant modality (commission error) (Schneider et al., 1984). Statistical analysis Statistical analysis was performed using Statistical Assistance System (SAS) statistical software (SAS Institute, 1990). The group comparison study was performed using analyses of variance (ANOVAs) with reaction time as a dependent variable. Reaction times were transformed (reciprocal transformation: v= 1000/reaction time) and only P values <0.05, obtained with both reaction time and transformation analyses, were considered significant (Milner, 1986). Post hoc analyses were performed with the Tukey Test (Glantz and Slinker, 1990). In order to take into account the role of the lesion volume in the right and left hemispheres, comparisons

5 Attention disorders in focal brain damage 195 between both patient groups were performed using analyses of covariance (ANCOVAs) with left and right lesion volumes as covariates (Stevens, 1992). Further analyses using multiple linear regressions were planned to assess (i) the role of the lesion's volume and location on attention disorders, (ii) the relationship between divided and focused attention, and (iii) the relationship between attention disorders and behavioural changes. Factors predicting poor performance were selected with a stepwise procedure according to a previously reported methodology (Godefroy et al., 1992). Dependent variables included three attention indexes: (i) the divided attention index obtained in detection tests; (ii) the crossmodal retardation index (crossmodal reaction time ipsimodal reaction time); (iii) the commission rate on Go/No-Go Tests. Multicollinearity was assessed using the variance inflation factor, and values lower than 10 were considered as reasonable (Glantz and Slinker, 1990). Results Detection tests In order to assess the ability to divide and shift attention, differences in reaction time to the different situations were evaluated using a three-factor ANOVA of subject ('frontal', 'posterior' and control groups), situation (unimodal, ipsimodal and crossmodal) and interstimulus intervals (1,4, 7 and 10 s). Effects of factors: subject [F(2) = , P < ]; situation [F(2) = , P < ]; interstimulus intervals [F(3) = 24.45, P < ] and the subjectxsituation interaction [F(4) = 24.43, P < ] were significant. Post hoc analysis revealed that patients in the 'frontal' group were slower to respond than both other groups (P < 0.05) and that both patient groups were slower than subjects in the control group (P < 0.05) (Table 3). Reaction times were longer in the crossmodal situation (P < 0.001) compared with the ipsimodal situation (the crossmodal retardation effect), and also in the ipsimodal situation (P < 0.001) compared with the unimodal situation (the ipsimodal retardation effect) (Table 3). The effect of factor interstimulus intervals was due to longer reaction times to stimuli preceded by a Is interval (P < 0.05) (Fig. 2). The subjectx situation interaction was attributable to higher ipsimodal retardation in the 'frontal' group (P < 0.01). The comparison of ipsimodal and crossmodal situations showed a significant subjectx situation X interval interaction [F(9) = 4.31, P < ] related to the higher crossmodal retardation in the 'frontal' group observed only for short interval (1 s) (Fig. 2). The ANCOVA provided similar results: reaction times were longer in the 'frontal' group [F(l) = 57.17, P< ], the ipsimodal retardation was higher in the 'frontal' group [subjectxsituation: F(2) = 18.84, P < ] and the crossmodal retardation for short intervals was higher in the 'frontal' group [subjectxsituationxinterval: F(6) = 2.58, P < 0.017]. 300 A Interstimulus interval Fig. 2 Mean reaction time [(visual reaction time+auditory reaction time)/2] of both patient groups in the three situations (unimodal, ipsimodal and crossmodal) of the detection tests according to the interstimulus interval. Closed diamonds = frontal unimodal; closed triangles = frontal ipsimodal; closed squares = frontal crossmodal; open diamond = posterior unimodal; closed triangles = posterior ipsimodal; closed squares = posterior crossmodal. These first analyses show that (i) reaction times were longer in both patient groups and in ipsimodal and crossmodal situations, and (ii) the response retardation in the ipsimodal and crossmodal situations was higher in the 'frontal' group, even when reaction times were adjusted for the lesion volume. The higher response retardation in the 'frontal' group in the ipsimodal situation indicates that the subjects were impaired when attention had to be allocated to two sensory modalities. Go/No-Go Tests Performance on Go/No-Go Tests was analysed as both mean reaction time and percentage of errors. The reaction time analysis used reaction times to both bimodal and Go/No-Go Tests in order to (i) assess the extent of the response retardation induced by the Go/No-Go decision, and (ii) evaluate the effect of modality shifting as a function of the task. The bimodal and Go/No-Go Tests used the same block of trials but differed with respect to the task. Thus, the overall response retardation on Go/No-Go Tests allowed us to evaluate whether the time required for the decision process was similar across groups. In addition, the Go/No-Go Tests required subjects to orient their attention to the attended sensory modality. The effect of modality shifting on reaction time may provide a means of evaluating whether attention was correctly oriented: focusing attention may attenuate the response retardation induced by modality shifting, i.e. the crossmodal retardation. In order to equalize the effect of

6 196 O. Godefroyet al. errors on reaction time across groups, reaction time following erroneous responses to No-Go stimuli were removed from the analysis. The analysis was achieved with a three-factor ANOVA of subject ('frontal', 'posterior' and control groups), test (bimodal, Go/No-Go) and situation (ipsimodal and crossmodal). It showed a significant effect of factors: subject [F(2) = , P < ], test [F(\) = , P< ] and situation [F(l)= 29.63, P< ]. Reaction times were longer in the 'frontal' group (P < 0.05) than in both other groups, and in both patient groups than in controls (P < 0.05), reaction times were also longer in Go/ No-Go Tests and in the crossmodal situation (Table 3). The overall reaction time increment in Go/No-Go Tests did not significantly differ across groups as shown by the nonsignificant subjectxtest interaction [F(2) = 0.21, P < 0.8]. The overall crossmodal retardation was more important in the 'frontal' group [subjectxsituation: F(2) = 6.45, P < 0.001]. The testxsituation interaction [F(l) = 4.86, P < 0.027] was significant and this was related to the lower crossmodal retardation in Go/No-Go Tests. In addition, the decrement in the crossmodal retardation in Go/No-Go Tests did not significantly differ according to the group [subject X situationxtest: F(2)= 1.1, P < 0.33]. The ANCOVA provided similar results: reaction times were longer in the 'frontal' group [F(l) = 49.25, P< ], in Go/No-Go Tests [F(l) =175, P< ], and in the crossmodal situation [F(l)=22.18, P< ]. The overall reaction time retardation in Go/No-Go Tests was similar in both groups [subjectxtest: F(\) =0.31, P < 0.57] and the overall crossmodal retardation was higher in the 'frontal' group [subjectxsituation: F(l)=7.04, P < 0.08]. Finally, the decreased crossmodal retardation in the Go/No-Go Tests did not significantly differ according to the group [subjectx situationxtest: F(l) =1.63, P < 0.2]. The Go/No-Go reaction time analysis indicates that the response retardation induced by the Go/No-Go decision did not significantly differ across groups, and that the effect of modality shifting mildly decreased in the Go/No-Go task. Furthermore, the overall crossmodal retardation was higher in the 'frontal' group, showing their higher vulnerability to the effect of modality shifting whatever the task. Focused attention was evaluated by the ability to suppress response to irrelevant stimuli. In order to evaluate the modality shift effect and the nature of errors, the analysis was performed with a three-factor ANOVA of subject ('frontal', 'posterior' and control groups), situation (ipsimodal and crossmodal) and error type (omission and commission) using the percentage of errors as the dependent variable. Heterogeneity of variance across groups dictated the use of a rank transformation of percentage of errors (Conover and Iman, 1981). It showed a significant effect of factors subject [F(2) = 17.37, P < ], errors being significantly higher in the 'frontal' group (P < 0.05), situation [F(\) = 16.99, P < ] and error type [F (1) = 109, P < ] and a significant subjectxerror type interaction (F = 15.08, P < ). Errors were more frequent in the 'frontal' group and in the crossmodal situation, and were mainly of the commission type (Table 3). The subjectxerror type interaction was related to the higher rate of commission errors in the 'frontal' group contrasting with the higher rate of omissions in the 'posterior' group (Table 3). The ANCOVA analysis provided similar results: the overall error rate was higher in the 'frontal' group [F(l) = 13.2, P < ] and in the crossmodal situation [F(l) = 5, P< 0.025]; errors were mainly of commission type [F (1) = 97, P < ] and commissions were more frequent in the 'frontal' group whereas omissions were more frequent in the 'posterior' group [subjectxerror type interaction: F(l) = 22.28, P < ]. Errors were more frequent in the 'frontal' group, in the crossmodal situation, and were mainly of the commission type. This result shows that the ability to suppress irrelevant responses was impaired in patients with frontal damage. According to the Signal Detection Theory approach (Green and Swets, 1966), two parameters influence the hits score: (i) the sensitivity, i.e. the ability to discriminate between signal and noise; (ii) the response criterion that refers to the level of the response threshold. Further analyses were performed in order to ascertain whether errors were related to the impairment of sensitivity or response criterion, or both. The analyses were performed using nonparametric indexes of sensitivity [P(I)] and criterion (B') (Grier, 1971) as dependent variables in a two-factor multivariate ANOVA of subject ('frontal', 'posterior' and control groups) and situation (ipsimodal and crossmodal). Multivariate planned comparisons were performed using contrast analysis (Stevens, 1992). It showed a significant effect of factor subject [F(2) = 3.59, P< 0.037] and situation [F(l) = 7.12, P<0.0\] and the subjectxsituation interaction approached significance [F(2) = 2.86, P < 0.069]. The effect of factor subject was attributable to the impairment of both indexes in the 'frontal' group [P(I), F(l) = 5.33, P < 0.026; B', F(l) = 5.82, P < 0.02]. The effect of factor situation was related to a decrement of sensitivity and criterion in the crossmodal situation (Table 3) and the subjectxsituation interaction was attributable to lower changes in both indexes according to the situation in the 'posterior' group [P(I), F(l) = 6.7, P < 0.013; B', F(l) = 6.91, P < 0.012] (Table 3). Thus, the Signal Detection Theory analysis shows that both sensitivity and response criterion were impaired in patients with frontal damage. In conclusion, the Go/No-Go Tests analyses show impaired performance in the 'frontal' group characterized by a higher sensitivity to the stimulus modality shifting and more frequent responses to irrelevant stimuli. This inability to reject irrelevant information fits the criterion of a deficit of focused attention (Schneider et al., 1984). The higher error rate in the 'frontal' group was related to the adoption of a more liberal response criterion and to decrement in sensitivity. Role of lesion location and volume The comparison and subsequent ANCOVA analyses of the three groups have shown that the 'frontal' group differed

7 Attention disorders in focal brain damage 197 from both other groups by a higher response retardation in the ipsimodal situation of the detection task and higher erroneous responses to irrelevant stimuli in the Go/No-Go task. This suggests that the presence of an anterior cerebral lesion may disrupt divided and focused attention. However, the damaged areas were variable within each patient group (Table 1) and this may account for the performance variability in patients (Godefroy, 1994). This between-subjects variability was used to evaluate the role of lesion location and size on the ability to divide and focus attention. Stepwise regression analyses were used to select the damaged areas and lesion size which predicted the slowing of perceptuomotor speed (unimodal reaction time), the divided attention index, the crossmodal retardation and the commission rate in the Go/No-Go task. Independent variables were age (years), education level, left and right lesion volume (cm 3 ) and presence of a lesion (0 = no; 1 = yes) in the following left and right areas: frontobasal (Brodmann areas 11 and 32), dorsolateral (Brodmann areas 8, 9, 10, 44, 45, 46 and 47), parietal and temporal lobes, thalamus and head of the caudate nucleus. The unimodal reaction time depended on age (r 2 = 0.16, P < 0.003) and on the presence of a lesion in (i) the left dorsolateral frontal cortex (r 2 = 0.21, P < 0.002), (ii) the right temporal lobe (r 2 = 0.08, P < 0.02) and (iii) caudate nucleus (^ = 0.035, P < 0.12). The index of divided attention depended on (i) the presence of a left lesion of the dorsolateral frontal cortex (r 2 = 0.495, P < ), (ii) on the right lesions volume (r 2 = 0.04, P < 0.079) and (iii) on age (r 2 = 0.04, P < 0.076). The crossmodal retardation in the Go/No-Go task depended on the presence of a lesion of the caudate nucleus (r 2 = 0.14, P < 0.015). The commission rate depended on the presence of a left lesion of the caudate nucleus (r 2 = 0.20, P < 0.003). These results indicate that (i) the slowing of perceptuomotor speed mainly depended on the presence of a lesion of the left dorsolateral frontal cortex, and age, (ii) the presence of divided attention deficit mainly depended on the presence of a lesion of the left dorsolateral frontal cortex, and (iii) commission rate mainly depended on the presence of a left lesion of the head of the caudate nucleus. Relationship between indexes of attention Attention capacity was evaluated using detection and Go/ No-Go tasks that provided various indexes of attention. The present analyses were planned to examine the relationship between these indexes. Stepwise regression analyses were used to select factors predicting the divided attention index, the crossmodal retardation and the commission rate in Go/ No-Go Tests. The lack of available data prevented prior selection of independent variables. This led to the inclusion of general factors (age, education level, unimodal reaction time and omission rate on Go/No-Go task) and attentional factors except those being evaluated [divided attention index (removed from the analysis of divided attention), commission 250i 20O 50 Divided attention index (ms) Commission rate Fig. 3 Divided attention index (ipsimodal reaction time - unimodal reaction time) computed from the detection tests and commission rate (per cent) on Go/No-Go tests: controls (filled circles) mean and 2 SDs. Three patients displayed a dissociated pattern: two had a selective impairment of the divided attention index (squares) and, the last one, of the commission rate (open circles). rate (removed from its analysis), crossmodal retardation in Go/No-Go Tests (removed from its analysis)] in the model. In order to evaluate the selectivity of attention disorder, we examined the individual divided attention index, crossmodal retardation and commission rate on the Go/No-Go task. Indexes deviating from the control-means by >2 SDs were considered to be impaired.,the regression analyses showed that the divided attention index was related to (i) unimodal reaction time (r 2 = 0.466, P < ), (ii) commission rate (r 2 = 0.076, P < 0.01) and (iii) crossmodal retardation in Go/No-Go Tests (r 2 = 0.075, f<0.01). The crossmodal retardation in Go/No-Go Tests was related to (i) commission rate (r 2 = 0.379, P < ) and (ii) divided attention index (r 2 = 0.146, P < 0.001). Finally, the commission rate was related to (i) crossmodal retardation in Go/No-Go Tests (r 2 = 0.379, P < ) and (ii) divided attention index (r 2 = 0.217, P < ). These analyses showed a close link between divided and focused attention indexes. However, examination of individual data showed that, first, one patient ('frontal' group) had impaired commission rate with normal divided attention index and that two patients (one in each patient group) displayed the reverse pattern (Fig. 3). The only patient in the 'posterior' group with a selective impairment of divided attention had an associated infarct of the left globus pallidus. Secondly, one patient ('frontal' group) had an impaired crossmodal retardation in Go/No-Go Tests contrasting with a normal commission rate and another patient ('frontal' group) displayed the reverse pattern (Fig. 4). These double dissociations observed in five patients indicate that divided

8 O. Godefroy et al. -20 Crossmodal retardation (ms) Commission rate Fig. 4 Crossmodal retardation = crossmodal reaction time - ipsimodal reaction time) and commission rate (percent) on Go/ No-Go tests: controls (circles) mean and 2 SDs; two patients displayed a dissociated pattern: one had a selective impairment of the crossmodal retardation (filled squares) and, the other one, of the commission rate (open squares). and focused attention may be selectively impaired and that both indexes computed from the Go/No-Go task (crossmodaj retardation and commission rate) may be selectively affected. Relationship between attention and behavioural changes These final analyses were conducted in order to examine the relationship between behavioural changes and attention. Factors related to behavioural changes (scores to aspontaneity and distractibility scales) were selected using stepwise regression analyses. The independent variables were age (years), education level, divided attention index, crossmodal retardation, omission and commission rates in Go/No-Go Tests. In order to examine whether the nature of the association between attention and behavioural changes differed according to the group, the predictive equations of each group were compared (Stevens, 1992). When the equation differed between groups (P < 0.1), the analysis was subsequently performed in each patient group. Aspontaneity was related to age (r 2 = 0.136, P < 0.09) and distractibility was related to (i) index of divided attention (r 2 = 0.06, P<0.12), (ii) commission rate (^ = 0.50, P < ) and (iii) crossmodal retardation in Go/No-Go Tests (r 2 = 0.08, P < 0.05); equations did not differ between groups. These analyses indicate that distractibility only was dependent on attention indexes. Discussion The group comparison study shows that patients with frontal damage (i) become increasingly slow to respond when attention has to be allocated to two sensory modalities, (ii) are more sensitive to the effect of stimulus modality shifting in both detection and Go/No-Go tasks, and (iii) exhibit more errors on Go/No-Go Tests that are related to the adoption of a more liberal criterion and to decrement in sensitivity. The transitive nature of attention prevents a direct assessment of its manifestations. This usually leads to the comparison of the performance of a simple task with that of a more complex version adding a step that is assumed to involve attention, as in the Stroop (1935) task. Following this line, we compared performance on unimodal and bimodal tests and on detection and Go/No-Go tasks. This led us to define three major markers of attention: (i) the divided attention index from the detection task; (ii) the crossmodal retardation; (iii) the commission rate in Go/No-Go Tests. The divided attention index was related to the performance decrement observed when targets to be detected arose from two sources. Unimodal and bimodal tests only differed with respect to the predictability of the sensory modality. Uncertainty about modality requires the allocation of attention to different modalities (Boulter, 1977). Thus, the response retardation observed in the ipsimodal situation may primarily reflect the division of attention across modalities, and the disproportionate retardation of the 'frontal' group supports the hypothesis of impaired divided attention. The response slowing in the crossmodal situation is likely to be due to attention shifting (LaBerge, 1973) and the higher crossmodal retardation of the 'frontal' group suggests a slowing of attention shifting between modalities. The Go/No-Go task was used to evaluate the ability to focus attention on the relevant source and its particularity depended upon the response criterion: it was based on stimulus modality and this offered the possibility to orient attention to one sensory modality. According to the Signal Detection Theory (Green and Swets, 1966), the present Go/ No-Go Tests were required to (i) orient attention to the relevant perceptual channel, (ii) compare the transduced stimulus with an internal representation and (iii) activate the motor response only if the accumulated comparisons exceeded the response threshold (Sperling, 1984). Thus, the focused attention phenomenon resulted from attention modulation that presumably operated at two different levels: (i) the orienting of attention to the relevant perceptual channel; (ii) the modulation of response threshold. The orienting of attention is likely to decrease the noise/signal ratio resulting in the increase in the sensitivity index (Green and Swets, 1966; Shiffrin and Grantham, 1974; Mangun et al., 1993). Another consequence of orienting attention to one source is the attenuation of the modality shift effect that was evaluated by the crossmodal retardation in Go/No-Go Tests. However, the crossmodal retardation effect is a sequential effect characterized by a small amplitude relative to reaction time, and this results in a poor reliability. Both the decrement in sensitivity and the persistence of high crossmodal retardation in Go/No-Go Tests were evident in the 'frontal' group suggesting that their ability to orient attention to the relevant

9 Attention disorders in focal brain damage 199 perceptual channel is impaired. The modulation of response threshold is also a critical parameter as it determines the suppression of response to irrelevant stimuli. Patients in the 'frontal' group adopted a more liberal criterion, which accounts for their higher commission rate and suggests that they are also impaired at the stage of the response selection; this is consistent with previous observations (Drewe, 1975; Salmaso and Denes, 1982; Leimkuhler and Mesulam, 1985). Therefore, both mechanisms that account for focused attention in the present Go/No-Go task, orienting of attention and response selection, were impaired in the 'frontal' group. The three major conclusions of the present study are that (i) divided and focused nonspatial attention may primarily depend on the dorsolateral prefrontal cortex, (ii) cerebral damage may selectively affect divided and focused attention and (iii) distractibility is closely related to both focused and divided attention. These points are discussed following the influence of extraneous factors on reaction time in patients with brain damage. The interpretation of reaction times in patients with brain damage is usually obscured by the possible role of extraneous factors, e.g. fatigability, training, motivation, vigilance level and reliability of reaction time measurement (Benton, 1986; Milner, 1986). In previous studies, we have shown that these factors do not account for the slowing of the patient's response (Godefroy et al., 1994a, b). In the present evaluation, we used tests that are easily administered to patients with brain damage using simple stimuli and motor responses. These tests did not involve complex cognitive processes and did not require high memory-load. Therefore, attentional disorders are unlikely to be due to either a decline in short term memory or an increase in attentional demands secondary to the inability to process complex tasks. Role of the frontal lobe in attentional control The present data indicate that patients with frontal damage are impaired on attentional tasks relative to patients with a lesion of the posterior cortex. Analysis of individual data replicated this main finding, with the exception of one patient in the 'posterior' group with an associated lesion of the globus pallidus who had impaired divided attention. This impairment could be due to the role of the globus pallidus in attentional control as previously suggested (Haaxma et al., 1993). Comparison of patients with posterior and anterior brain damage was complicated by the higher volume of rightsided lesions in the 'frontal' group. In order to take into account this prominence of the right lesion size in the anterior group, we adjusted both groups' comparisons for lesion size and included it as independent variable in the regression analyses. This procedure left the main results unchanged suggesting that the main factor that accounts for attention disorders was the presence of a lesion in the anterior part of the brain. The ability to monitor targets arising from two sources was dependent on the presence of a lesion of the dorsolateral prefrontal cortex; this is consistent with the role of the dorsolateral prefrontal cortex in sensory gating (Knight, 1991) and attentional filtering (Roland, 1982; Mesulam, 1985). The impairment of attention regulation that was observed in the simple detection task supports the idea that the role of the prefrontal cortex is not restricted to the information uptake or to the output selection required in complex cognitive tasks, but that it is already operating at the stage of simple perceptuomotor processes. The number of erroneous responses to irrelevant stimuli in Go/No-Go tasks was dependent on the presence of a lesion of the head of the caudate nucleus. This suggests that the accuracy of response selection depends on the integrity of the caudate nucleus. A previous study that assessed patients with striatal infarcts also supported the involvement of the caudate nucleus in the selection of the motor programme (Godefroy et al., 1992). Both results support the role of the caudate nucleus at the stage of response selection. Previous investigations have suggested that the side of the lesion may influence the extent of attentional deficit and slowing of reaction time (Mesulam, 1985; Tartaglione et al., 1986; Posner and Rafal, 1987). The prominent role of the right hemisphere has been demonstrated in spatial attention, whereas nonspatial attentional control may depend upon the contribution of both hemispheres (Vendrell et al., 1995; Godefroy and Rousseaux, 1996). Studies comparing the effect of anterior versus posterior loci of brain damage on attention have seldom evaluated the lesion size according to the side. The effect of the lesion size and side appears complex and dependent on the task considered. In the present study, the role of right-sided lesions was prominent for simple unimodal reaction time; this is consistent with the results from Boiler et al. (1970) and Tartaglione et al. (1986). Conversely, the proportion of variance of divided and focused attention indexes accounted for by the right lesion size was lower, contrasting with the prominent role of left-sided lesions. This is consistent with the findings of Dee and Van Allen (1973) that showed a more marked response retardation as a function of task complexity in patients with left-sided brain damage only. Therefore, the contribution of both hemispheres to nonspatial attentional control may depend on the task and probably on the processing stage considered. Behavioural changes and attention In this study, we demonstrated a relationship between distractibility and attention indexes that is unlikely to be due to anatomical contiguity as it was not group-dependent. This result supports the idea that attention disorders may have a functional counterpart that is clinically assessable, consistent with classical views in neuropsychology. For example, Hecaen (1964) highlighted the distractibility, and the necessity for repeated questioning, to obtain a response in frontal lobe patients that he attributed to a disorder of attention. Since the publication of the case of Phineas Gage (Harlow, 1868), it is well known that frontal damage may alter the regulation

10 200 O. Godefroy et al. of behaviour and emotion. Behavioural abnormalities are frequently attributed to the involvement of the paralimbic components in the frontal lobe that are thought to be of crucial importance for channelling drive and emotion (Mesulam, 1986; Fuster, 1989). However, our results are consistent with the idea that some behavioural abnormalities may be secondary, at least in part, to attention deficit. Relationship between varieties of attention The present study shows that two varieties of attention may be impaired selectively by the presence of brain damage: three patients displayed a selective deficit of divided or focused attention. This dissociated pattern suggests that mechanisms that support divided and focused attention are different. Moreover, within the same Go/No-Go task, two measures of performance level were affected selectively: (i) the crossmodal retardation effect that is related to the orienting of attention; (ii) the number of commission errors, related to the decision process. The latter result is consistent with the study of Vendrell et al. (1995) that shows that the general cognitive speed and modulation of verbal responses on the Stroop Test depend on different frontal areas. Such observations run counter to theories that propose a unique impairment of attention regulation following frontal damage. Conversely, Vendrell et al. (1995) also suggest that different areas within the anterior part of the brain are involved in various attentional processes that concern simple perceptuomotor processes, orientation of attention and the modulation of parameters regulating decision processes. Furthermore, evidence for different attentional mechanisms argues against a unitary conception of attention. The conception of attention as a unitary and separable system is probably relevant to arousal and alertness phenomena that may be supported by specific neuronal structures (Mesulam, 1985; Stuss and Benson, 1986). However, our data, as well as those of other studies on spatial attention, provide evidence for attentional modulations operating at different levels within the same task and suggest that they depend on different anatomical systems (for review, see Allport, 1993). Recent advances in visual processing have shown that both spatial and categorical processes operate in parallel and that they are supported by different anatomical pathways (Mishkin et al., 1983). This contrasts with the earlier prevailing view of attentional selection prominently operating on spatial processes (Schneider et al, 1984; Allport, 1993). More recently, Posner and Rafal (1987) suggested that visual spatial attention involves different mechanisms that depend on several, closely interrelated, anatomical structures. These data argue against the unitary nature of attention and support the idea that attention depends on several mechanisms that are probably task-dependent which may be externally perceived as unique. Therefore, reference to attention as an undifferentiated, general-purpose mechanism may be irrelevant. The conceptual need for a general regulation of the many specific-purpose cognitive systems in order to produce coherent actions, has led some authors to postulate the existence of a unitary central executive or supervisory or anterior attentional system (Norman and Shallice, 1986; Baddeley, 1986; Posner and Petersen, 1990). The model developed by Norman and Shallice (1986) provides theoretical accounts of Luria's approach, specifically the prominent impairment of complex or novel tasks in frontal lobe pathology (Luria, 1978). Moreover, this model accurately predicts numerous cognitive impairments specific to frontal lobe damage (Shallice, 1982, 1988; Shallice and Burgess, 1991a, b). Despite the clear advance provided by this model, evidence for different modulation mechanisms supports the idea that the supervisory system involves different regulation systems as suggested by Shallice (1994). This view is consistent with dissociated deficits within executive processes as demonstrated by Eslinger and Damasio (1985) and Shallice and Burgess (1991a, b) and also with functional imaging studies that suggest that different processes regulate voluntary behaviour (Roland, 1982). Considering the evidence of multiple levels of control processes and of attentional modulation mechanisms, further studies that examine the nature of regulation systems and subsequent implications for the cognitive architecture are required as suggested by Shallice (1994). Conclusion In conclusion, our results lend support to the view that the prefrontal cortex is the critical node of networks for nonspatial attentional controls that are exerted during both simple and strategically organized cognitive processes. In addition, our study provides data that suggest that attention disorders may have a behavioural counterpart. Finally, evidence for a selective deficit of focused and divided attention suggests that attention may depend upon various, and probably different, mechanisms. This may lead to a fractionation of the evaluation of attention and preclude reference to attention as an unspecified causal mechanism. From this point of view, further studies addressing the nature of the interactions between various attentional mechanisms and other cognitive processes, such as those involved in planning or memory, are required. Acknowledgements We wish to thank Maryline Cabaret, Violaine Petit and Brigitte Debachy for their technical assistance in the neuropsychological assessment; Jean Pierre Pruvo for the neuroradiological evaluation; Didier Leys, Francis Lesoin and Jean-Paul Lejeune who allowed us to investigate some of their patients; Michele Brouchon and Marc Jeannerod for their helpful comments, and Patricia McBride for reviewing the manuscript. This study has been supported by grants from the Direction de la Recherche et des Etudes Doctorales

11 Attention disorders in focal brain damage 201 (No ), Faculte de M6decine de Lille and from the Laboratoires Schering SA. References Alivisatos B. The role of the frontal cortex in the use of advance information in a mental rotation paradigm. Neuropsychologia 1992; 30: Allport A. Attention and control: have we been asking the wrong questions? A critical review of twenty-five years. In: Meyer DE, Kornblum S, editors. Attention and performance XIV. Cambridge (MA): MIT Press, 1993: Baddeley AD. Working memory. Oxford: Clarendon Press, Benton A. Reaction time in brain disease: some reflections. Cortex 1986; 22: Boiler F, Howes D, Patten DH. A behavioral evaluation of brainscan estimates of lesion size. Neurology 1970; 20: Boulter LR. Attention and reaction times to signals of uncertain modality. J Exp Psychol Hum Percept Perform 1977; 3: Broadbent DE. Pergamon, Perception and communication. Oxford: Conover WJ, Iman RL. Rank transformation as a bridge between parametric and nonparametric statistics. Am Stat 1981; 35: Dee HL, Van Allen MW. Speed of decision-making processes in patients with unilateral cerebral disease. Arch Neurol 1973; 28: Drewe EA. Go/no-go learning after frontal lobe lesions in humans. Cortex 1975; 11: Eslinger PJ, Damasio AR. Severe disturbance of higher cognition after bilateral frontal lobe ablation: patient EVR. Neurology 1985; 35: 1731^1. Folstein MF, Folstein SE, McHugh PR. 'Mini-mental state': a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: Fuster JM. The prefrontal cortex. 2nd ed. New York: Raven Press, Glantz SA, Slinker BK. Primer of applied regression and analysis of variance. New York: McGraw-Hill, Godefroy O. Frontal lobe dysfunction in vascular dementia. In: Leys D, Scheltens P, editors. Vascular dementia. Dordrecht: ICG Publications, 1994: Godefroy O, Rousseaux M. Divided and focused attention in patients with lesion of the prefrontal cortex. Brain Cogn In press. Godefroy O, Rousseaux M, Leys D, Destee A, Scheltens Ph, Pruvo JP. Frontal lobe dysfunction in unilateral lenticulostriate infarcts. Arch Neurol 1992; 49: Godefroy O, Cabaret M, Rousseaux M. Vigilance and effects of fatigability, practice and motivation on simple reaction time tests in patients with lesion of the frontal lobe. Neuropsychologia 1994a; 32: Godefroy O, Lhullier C, Rousseaux M. Reliability of reaction time measurements in brain-damaged patients. J Neurol Sci 1994b; 126: Green DM, Swets JA. Signal detection theory and psychophysics. New York: J Wiley, Grier JB. Nonparametric indexes for sensitivity and bias: computing formulas. Psychol Bull 1971; 75: Haaxma R, Robbins TW, James M, Brouwer WH, Colebatch JG, Marsden CD. Neurobehavioural changes in a patient with bilateral lesions of the globus pallidus. Behav Neurol 1993; 6: Harlow JM. Passage of an iron rod through the head. Mass M Soc Bost 1868; 2: 327^*6. Hecaen H. Mental symptoms associated with tumors of the frontal lobe. In: Warren JM, Akert F, editors. The frontal granular cortex and behavior. New York: McGraw-Hill, 1964: Knight RT. Evoked potential studies of attention capacity in human frontal lobe lesions. In: Levin HS, Eisenberg HM, Benton AL, editors. Frontal lobe function and dysfunction. New York: Oxford University Press, 1991: LaBerge D. Identification of two components of the time to switch attention: a test of a serial and a parallel model of attention. In: Komblum S, editor. Attention and performance IV. New York: Academic Press, 1973: Leclercq M, Chipp P, Blandel A, Bruyere R, Coyette F, Frederix M, et al. A simple reaction times set. European Symposium on microcomputers in neuropsychological assessment and rehabilitation; Strasbourg (France), Leimkuhler ME, Mesulam MM. Reversible go/no-go deficits in a case of frontal lobe tumor. Ann Neurol 1985; 18: Luria AR. Les fonctions corticales sup^rieures de l'homme. Paris: Presse Universitaire de France, Mangun GR, Hillyard SA, Luck SJ. Electrocortical substrates of visual selective attention. In: Meyer DE, Komblum S, editors. Attention and performance XIV. Cambridge (MA): MIT Press, 1993: Mesulam MM. Principles of behavioral neurology. Philadelphia: FA. Davis, Mesulam MM. Frontal cortex and behavior. Ann Neurol 1986; 19: Milner AD. Chronometric analysis in neuropsychology [review]. Neuropsychologia 1986; 24: Mishkin M, Ungerleider LG, Macko KA. Object vision and spatial vision: two cortical pathways. Trends Neurosci 1983; 6: Norman D, Shallice T. Attention to action: willed and automatic control of behaviour. In: Davidson RJ, Schwartz GE, Shapiro D, editors. Consciousness and self-regulation, vol. 4. New York: Plenum, Perret E. The left frontal lobe of man and the suppression of habitual responses in verbal categorical behaviour. Neuropsychologia 1974; 12: Phillips S, Sangalang V, Stems G. Basal forebrain infarction: a clinicopathologic correlation. Arch Neurol 1987; 44: