Action sequencing deficit following frontal lobe lesion

Neurocase (2002) Vol. 8, pp. 88–99

© Oxford University Press 2002

Action Sequencing Deficit Following Frontal Lobe Lesion Sergio Zanini1, Raffaella I. Rumiati1 and Tim Shallice1,2 1Programme

in Neuroscience, Scuola Internazionale Superiore di Studi Avanzati, Trieste, Italy and 2Institute of Cognitive Neuroscience, University College London, London, UK

Abstract Frontal lobe patients carried out temporal sequencing tasks related to actions that differed in terms of their abstractness using both verbal and pictorial presentations. A generalized impairment was found: neither a type of action effect nor a modality of item presentation effect was present. The patients also carried out a corresponding action production task and produced actions quickly and without errors. The frontal lobe patients were also spared in generating verbal descriptions of actions: they were as accurate as normal controls both in terms of the details reported and in maintaining the temporal sequence. It has been argued that the difficulty in processing the temporal dimensions of actions following frontal lobe lesions is due to some form of disruption of the action representation. However, no action representational deficits were present in our frontal lobe patients. Thus, they cannot account for our findings. On the contrary, we suggest that the action sequencing deficit was a consequence of the difficulties patients experienced in rejecting wrong alternatives presented by the stimulus situation.

Introduction One can consider the organization of human actions on a variety of levels. Schmidt (1975) and Arbib (1985) have argued that individual subcomponents of a larger skill unit can be treated as ‘motor response schemas’ formed by abstractions over movements. So during prehension there would be such schemas for ‘reaching’, ‘pre-shaping’, ‘enclosing’ and so on. At a much higher level are scripts (Schank and Abelson, 1977) and the memory organization packets (MOPs) of Schank (1982). They are held to represent the sequence of higher-level operations involved in a familiar activity like ‘going to a restaurant’ or ‘going to the doctor’. Between the two in level are activities such as ‘making a cup of coffee’, ‘brushing one’s teeth’, ‘starting a car’, where the individual subactions appear to be controlled by discrete operations such as ‘stir the coffee’ (in making a cup of coffee) which are in turn themselves very close to the lowestlevel motor response schemas. This third domain, which we call lower-level schemas (or schema-type actions), is that of simpler actions where each single step is typically carried out in a highly specific manner in order to reach the action goal. Normally when one prepares orange juice, say, the orange is cut in one particular manner, namely by sawing the orange and not by pushing down the knife as is done when one is cutting butter. By contrast, higher-level representations such as scripts or MOPs do not necessarily relate directly to effector systems. Thus, the step of ‘paying the bill’ in the ‘going to the restaurant’ script may be realized by making

out a cheque, counting out notes or signing a credit card slip. Moreover, at this higher level, operations are typically interleaved by many other activities such as visiting a shop when going to the doctor. This can also happen when brushing one’s teeth, say, but it is unusual and often counterproductive. Thus, the lower-level schemas and higher-level scripts differ on a number of dimensions [see Cooper and Shallice (2000) for a discussion]. Frontal patients have been reported to experience impairments in action processing at both the schema and the script levels. In particular, after Luria’s (1966) pioneering studies, several single-case studies have been carried out by Schwartz et al. (1991, 1995, 1998) and by Humphreys and Forde (1998), where attention has been mainly directed to deficits in the production of actions of the schema type following extensive lesions involving the frontal lobe. In this paper, we will, however, focus our attention on a second group of investigations (Sirigu et al., 1995, 1996, 1998; Partiot et al., 1996; Crozier et al., 1999) which have consistently shown the pre-frontal cortex to be involved at a higher level of cognitive control in the processing of the temporal structure of script actions. Sirigu et al. (1995) compared pre-frontal patients with posterior patients and normal controls. Several striking findings were reported. In a script generation task—the evoking of as many action steps as subjects could think of—the three groups reported approximately the same number of script steps and followed

Correspondence to: S. Zanini, Cognitive Neuroscience Sector, SISSA, Via Beirut 2–4, 34014 Trieste, Italy. E-mail: [email protected] or T. Shallice, Institute of Cognitive Neuroscience, University College London, Alexandra House, 17 Queen Square, London WC1N 3AR, UK. E-mail: [email protected]

Action sequencing deficit 89

the intrinsic temporal order of the action. However, only the pre-frontal patients made early closures of the scripts and made sequence errors which violated the temporal structure of the scripts. Moreover, the pre-frontal patients produced more errors in sequencing novel and non-routine scripts compared with routine ones. Sirigu et al. interpreted these data in the light of Grafman’s (1989) theory of managerial knowledge units (MKUs) which are defined as ‘single units of memory representing thematic and temporal aspects of event series...’ (Grafman, 1994). MKUs can be considered closely related concepts to scripts and MOPs. Pre-frontal patients evoked as many script steps as posterior patients and normal controls. Thus, basic representation of the scripts was held to be stored in other cortical areas in association with lexical and semantic information. By contrast, lesions to pre-frontal cortex were suggested to damage aspects of scripts such as temporal sequence, boundaries and priority sets, which the authors labelled as aspects of the script ‘grammar’ for logical and temporal order. Two comments can be made. First, it is argued that the basic representation of the script is spared after pre-frontal damage, but that the MKUs are not. The theory claims that the MKU is characterized by a functional network that assembles basic script information into the correct temporal syntax. In order to corroborate this theory, it would be important to verify the independence of the MKU from basic knowledge by comparing pre-frontal patients with posterior ones having basic semantic knowledge deficits, e.g. in semantic dementia. In fact, the posterior group, in this study, did not differ from the normal controls. No one has yet reported the contrasting dissociation of a disorder of script ‘grammar’, due to a pre-frontal lesion, compared with a ‘semantic’ deficit due to a posterior lesion. The second comment concerns the claim that these findings provide some support for MKU theory. Subjects had to give the steps involved in three situations: going to work, planning a trip to Mexico and starting up a beauty salon business. Only the first would qualify as just a script. The second and particularly the third contain major problem-solving elements, and these may be the cause of any frontal impairment, particularly in the generation task. In a following paper, Sirigu et al. (1996) investigated the same patients reported in Sirigu et al. (1995) on both script sequencing and script sorting tasks. They replicated the earlier findings on sequencing and also found a selective impairment in boundary maintenance of scripts in the prefrontal population. The patients tended to mix scripts together, to fuse them, and to include distracters into legitimate scripts or to create new implausible scripts with distracters only. The authors interpreted these findings by slightly modifying their previous (Sirigu et al., 1995) theoretical position. They suggested that sequence information for scripts can be encoded in two ways: the first by posterior cortex associative networks that store the coarse temporal contiguity between an initial state, an action and a final state, as well as semantic associations between script elements; the second by a frontal

network that can store any sequential patterns of activation. Thus, they claimed that the pre-frontal cortex ‘stores an arbitrary sequence using the consequence of the action as a binding agent between context and action, instead of a temporal correlation’ (Sirigu et al., 1996, p. 308). Therefore, the authors suggested that the finer temporal structure of scripts is automatically shaped when pre-frontal areas trigger script element activation in a top-down manner. In the authors’ view, this theory could account for frontal patients’ script processing performances characterized by the maintenance of a general chronological organization of scripts (due to intact posterior cortex representation), interrupted by frequent steps back in time (due to loss of more fine-grained pre-frontally stored information). This interpretation was not specifically supported by neuropsychological evidence, as no assessment of working memory was carried out. In a recent paper, Sirigu et al. (1998) carried out further investigations on the processing of sequences following frontal lobe lesions. They compared four Broca’s aphasics with lesions not involving the pre-frontal cortex and four pre-frontal patients with lesions not affecting Broca’s area. The Broca’s aphasics were able to sort and sequence action scripts but failed on a sentence ordering task. The pre-frontal patients were able to sequence parts of sentences correctly, respecting given syntactic cues, and to sort scripts but were impaired in ordering script elements. This double dissociation between sentence syntax and story grammar sequencing was interpreted as resulting from the independence of the underlying cognitive processes. The process of ordering, typically attributed to frontal lobes, was believed to be specific to the type of item to be ordered. These authors claimed that it is unlikely that ordering depends ‘on a supramodal processor, but [it] is rather a function of the type of underlying knowledge structure to be processed’ (Sirigu et al., 1998, p. 776). Two neuroimaging studies addressed script processing in normal subjects (Partiot et al., 1996; Crozier et al., 1999). In the Partiot et al. (1996) positron emission tomography study, normal subjects were expected to judge whether a given script step belonged to a certain script (content aspects of the script) and to judge whether a given script step occurred before or after another one (formal temporal aspects of the script), with a font discrimination task as the baseline condition. They found activations of the right frontal lobe, the left superior temporal gyrus and the middle temporal gyrus bilaterally during the first task. In contrast, the left frontal lobe, the left anterior cingulate and the anterior part of the left superior temporal gyrus were more active during the second task. The authors concluded that temporal ordering of scripts and determining whether an event belongs to a particular script seem to be processed by distinctive neuronal networks. In the Crozier et al. (1999) functional magnetic resonance imaging study, the authors tried to confirm Sirigu et al.’s (1998) findings. They reported a major activation of right and left middle frontal gyri, the left supplementary motor

90 S. Zanini, R. I. Rumiati and T. Shallice

area and the left angular gyrus in a script sequencing task compared with a sentence sequencing task. They interpreted this finding by claiming that the regions activated in this condition contain neural circuits involved in action planning and specifically in their temporal ordering when representations from long-term memory are recalled. They made the highly speculative suggestion that these areas are recruited when an event sequence representation is activated, in analogy to Bottini et al.’s (1994) evidence of right middle frontal activation during metaphor plausibility judgement tasks. In fact, the study failed to show a clear-cut double dissociation. By contrast to the several areas that were more activated during script processing, no areas were more active during the syntactic task. Therefore, the findings basically corroborate a pre-frontal involvement in script processing. As can be seen, Sirigu et al. have provided extensive evidence that the frontal lobes are heavily involved in processing the temporal structure of scripts. However, the theoretical accounts provided are very speculative. No strong evidence has been obtained for either the script syntax representation or the working memory hypotheses. In addition, these group studies on temporal processing of actions in frontal lobe patients have concentrated on one type of action only (i.e. script-level actions) and one type of modality only (verbal presentation of stimuli) (Sirigu et al., 1995, 1996, 1998). Therefore, the present study had the following aims: (1) to assess, by analogy with Sirigu et al.’s studies, whether the ordering of different levels of actions, namely one pertaining to the schema level (Schimdt, 1975; Norman and Shallice, 1986; Cooper and Shallice, 2000) and more abstract ones pertaining to the script level (Schank and Abelson, 1977) are processed differently in frontal lobe patients and to assess whether different modalities of item presentation (verbal and pictorial) have an effect on access to action knowledge; (2) to assess whether the subcomponents of action could be effectively ordered in other types of task; (3) to assess whether action sequencing impairments correlate with more basic sequencing deficits in frontal lobe patients.

Section 1. Action sequencing Experiment 1 (conditions a, b, c, d) Four different action sequencing tasks were given to the experimental subjects. Two sets of schema-type actions (e.g. ‘preparing orange juice’) (experiments 1a and 1b) and two sets of script-type actions (e.g. ‘having a meal at the restaurant’) (experiments 1c and 1d) were used. Each had to be carried out in two different conditions, using pictures (experiments 1a and 1c) and verbal descriptions (experiments 1b and 1d). Ten schema-type actions and 10 script-type actions were used (see Appendices 1 and 2). The selected actions were highly familiar to all patients. Only salient action steps—steps belonging to what we will call an ‘ordinate’ level (Schank and Abelson, 1977)—were included in the sequence (e.g. for schemas—‘opening a tin of tuna fish’: holding the tin opener, cutting the lid, bending back the lid, lifting the tuna on to a plate with a fork; for scripts—‘going to the doctor’: sitting in the waiting room, entering the doctor’s office, undressing, lying on the bed, dressing, leaving the doctor’s office). The same steps were presented in the verbal and pictorial modalities. We asked seven normal subjects, not involved in the study, to check whether the action sequences included subordinate- rather than ordinate-level steps (e.g. in the script ‘going to the doctor’, the step ‘undressing’ might be described also including subordinate actions such as ‘unbuttoning the shirt’ or ‘taking shoes off’). They all agreed that the sequences did not do so. Using this criterion, the schema sequences ranged from four to six steps while the script sequences ranged from five to eight steps. A complete balancing of action sequence lengths between schemas and scripts was not possible due to the different ranges of sequence lengths that naturally occur. Cards, each representing a single action step, were placed in a pseudo-random order on the table in front of the subject. All the cards had to be changed to obtain the correct solution. All subjects were given the same order for any given set of stimuli. No time limits were imposed. They were allowed to rearrange the cards as many times as they wished. To be scored correct, the sequence had to be eventually correct. The patients were tested in the four experimental conditions in four separate testing sessions.

Results Subjects Nine consecutive right-handed patients with cerebral lesions involving frontal lobes (in three cases the lesion extended to the temporal lobe) (Fig. 1) were tested. A set of baseline neuropsychological tests was administered to assess general intelligence, language functions, attention, praxis abilities, executive functions, memory and visuospatial abilities (see Table 1). Nine age-matched normal controls [mean age 60.11 years, standard deviation (SD) 12.11] were compared with the frontal lobe patients.

The patient group scored far below the control mean on all action sequencing tasks (see Fig. 2). In experiment 1a (schema–pictorial) their mean score on all correct sequence rearrangements was 33.3% (SD ⫽ 23.4), highly significantly less than the normal mean of 95.6% (SD ⫽ 7.3) (t ⫽ –7.91, P ⬍ 0.0001). In experiment 1b (schema–verbal), the patients rearranged 37.8% (SD ⫽ 28.6) sequences correctly while the normal controls scored 90% (SD ⫽ 13.2) correct, highly significantly more (t ⫽ –4.76, P ⬍ 0.0001). In experiment 1c (script–pictorial), the frontal lobe group again had a pathological performance: their mean score of 23.3%

Action sequencing deficit 91

Fig. 1. Lesion diagrams of each of the nine patients (right and left sides are inverted such as in computed tomography scans).

(SD ⫽ 28.7) was significantly less than the normal mean of 80% correct (SD ⫽ 22.3) (t ⫽ –4.54, P ⬍ 0.0001). In experiment 1d (script–verbal), the patients scored 24.4% correct (SD ⫽ 27.9), significantly worse than the control

subjects who scored 96.7% (SD ⫽ 5) correct (t ⫽ –7.4, P ⬍ 0.0001). An inter-test correlation analysis revealed that the script– pictorial and schema–pictorial tests correlated highly with

92 S. Zanini, R. I. Rumiati and T. Shallice Table 1. Neuropsychological assessment Patient 1

2

3

4

5

6

7

8

9

Age (years) Lesion site Brodmann areas

75 LF 6,44,45,46

48 LF 6,24,45

74 LF(T) 6,9,22,41, 42,44,45,47

38 BF 8,9,10, 24,32,45,46

75 LF(T) 6,9,10, 22,44,45,46

64 RF 8,9,10,24, 32,44,45,46

27 BF 4,6,24,32

Aetiology Intelligence WAIS IQ Language AAT Attention Attention matricese Praxis Ideomotor apraxia Ideational apraxia Buccofacial apraxia Constructional apraxia Executive functions Weigl (No. categories) WCST (No. categories) Perserverative errors (%) Brixton (scaled score) Reversed digit span Memory Digit span Story recall Corsi test Visuospatial ability VOSP screening VOSP object decision BORB (T8) BORB (T12)

Vascular

Vascular

Vascular

Traumatic

Vascular

Vascular

Traumatic

64 BF 9,10,12, 24,32,44, 45,46 Traumatic

59 BF(T) 6,9,10,11, 12,21,22,24, 31,38,45,46 Vascular

86

92

95

93

87

66

76

86

86

Broca

Broca

Broca

Not aphasic

Broca

Anomic

Anomic

Not aphasic Not aphasic

28/60

35/60a

20/60a

48/60

29/60

17/60a

57/60

44/60

44/60

41/72b 14/14 10/20a 6/14a

60/72 14/14 15/20a 12/14

44/72b 14/14 11/20a 9/14a

68/72 14/14 19/20a 10/14a

61/72 14/14 14/20a 3/14a

38/72b 14/14 18/20a 2/14a

64/72 14/14 17/20a 10/14a

61/72 14/14 20/20 9/14a

62/72 14/14 17/20a 14/14

1b

1b

1b

2

2

1b

2

1b

1b

1a 48c nt nt

2a 61c nt nt

1a 32 nt nt

2a 63c 3 3a

2a 76c 1 nt

1a 77c 1 2a

2a 32 nt 5

1a 79c 1 3a

1a 100c 2 2a

9/24d nt 2a

6/24d nt 5

11/24d nt 4

6 3.3/16a 5

11/24d 6.6/16 4

4a 3/16a 4

5 11.6/16 6

5 4.2/16a 4

6 5.3/16a 4

20/20 17/20 21/25 25/30

16/20a 17/20 23/25 27/30

20/20 15/20 22/25 30/30

20/20 16/20a 24/25 29/30

16/20a 11/20a 21/25 26/30

20/20 11/20a 18/25 21/30a

20/20 19/20 24/25 26/30

20/20 16/20 22/25 26/30

20/20 17/20 24/25 30/30

aPerformance falling 1.5 or more standard deviations bPerformance falling below cut-off. cPerformance falling above the pathological cut-off. dProbe recognition (six words version). eSpinnler and Tognoni, 1987.

below normal mean.

L, left; R, right; B, bilateral; F, frontal; T, temporal; nt, not tested; WAIS, Wechsler Adult Intelligence Scale; AAT, Aachen Aphasia Test; VOSP, Visual Object and Space Perception battery; BORB (T8), Birmingham Object Recognition Battery foreshortened match; BORB (T12), association match test; WCST, Wisconsin Card Sorting Test.

each other (Spearman’s rho ⫽ –0.76, P ⬍ 0.02), as did the script–verbal and schema–verbal tests (Spearman’s rho ⫽ –0.78, P ⬍ 0.02). The script–pictorial test sequencing test correlated (Spearman’s rho ⫽ –0.69, P ⬍ 0.04) with the buccofacial apraxia test, as did the script–verbal (Spearman’s rho ⫽ –0.76, P ⬍ 0.02) and the schema– verbal tests (Spearman’s rho ⫽ –0.79, P ⬍ 0.02). One possibility is that these correlations come from anatomical contiguity. The other significant correlation of the sequencing tests with a neuropsychological baseline test was the script–pictorial test with the Corsi block test (Spearman’s rho ⫽ –0.68, P ⬍ 0.04). The correlations of the Corsi block test with the other three sequencing tasks were far from significant (all P ⬎ 0.2). We further analysed the performances of the patients and controls in order to reveal whether there was an action-type or modality-type effect, namely whether they differed in

sequencing actions of different types or actions presented with different stimuli. An ANOVA for repeated measures with arc-sine transformation of raw data was run to test whether the subjects’ performances differed in respect to the type of action (schema versus script) or type of stimulus (verbal versus pictorial). A 2 ⫻ 2 ⫻ 2 design was set: type of action (schema versus script) ⫻ modality (verbal versus pictorial)⫻group (frontal patients versus control subjects). The main effect of group was significant: the patients’ performance was lower than that of the control subjects (F1,16 ⫽ 53.1, P ⬍ 0.0001). The main effect of type of action was also significant (F1,16 ⫽ 9.01, P ⬍ 0.01). Overall, the subjects made more errors in sequencing script actions than schema actions. No difference between verbal and pictorial presentation of stimuli was found (F1,16 ⫽ 0.53, P ⬎ 0.4). No interactions were found. One possibility is that the effect of type of action could

Action sequencing deficit 93 Table 2. Comparison of schema- and script-type actions controlled for sequence lengths expressed as a percentage of the total Patients

Total correct rearrangements Schema 5 versus script 5 Schema 6 versus script 6 Schema (5⫹6)/2 versus script (5⫹6)/2 Correct placement within action sequence Schema 5 versus script 5 Schema 6 versus script 6 Schema (5⫹6)/2 versus script (5⫹6)/2

Controls

Schema

Script

P

Schema

Script

P

26.8 29.7 28.2

23.0 25.4 24.2

ns ns ns

92.9 94.4 93.5

89.8 92.0 90.9

ns ns ns

43.2 55.6 49.4

42.6 52.0 47.3

ns ns ns

95.8 97.2 96.5

93.9 94.0 94.0

ns ns ns

Schema/script 5, schema/script-type actions of five steps; schema/script 6, schema/script-type actions of six steps; schema/script (5⫹6)/2, averaged score of schema/script-type actions of five and six steps; ns, not significant.

just be a result of the difference in length of the action step sequences between the schema and the script stimuli. For the former stimuli, sequences were of four to six steps whilst for the latter they ranged from five to eight. Correlation analyses between the length of action sequences and the number of correctly rearranged sequences of that length proved significant only in the patient group (Spearman’s rho ⫽ –0.41, P ⬍ 0.01). In order to assess the possibility that the overall type of action effect came from the average length difference between the groups, we compared the performance of the patients and the control subjects on schema versus script actions of the same length. The performances on schema and script actions of five and six steps were compared. In addition, the same comparisons were made on the percentage of correctly positioned action steps within action sequences. No significant difference was found for either comparison (see Table 2). Thus, the patients and the control subjects both had comparable performances for schema and script sequences when length was controlled. Thus, the performance of the patients and the control subjects on action sequencing tasks (experiments 1a, 1b, 1c, 1d) differed between schema and script actions simply because the scripttype action sequences were overall somewhat longer.

Error analysis We further analysed action sequences in order to determine whether there was a particular pattern of error occurrence in the sequences, contrasting the initial, the central and the last steps. An ANOVA with a 3 ⫻ 2 design was carried out: position (first versus central versus last)⫻group (frontal patients versus controls), with data being transformed using the arc-sine procedure. The main effects of position (F2,156 ⫽ 26.04, P ⬍ 0.0001) and of group (F1,78 ⫽ 318.02, P ⬍ 0.0001) and the interaction position⫻group (F2,156 ⫽ 4.2, P ⬍ 0.017) were all significant. Overall, the central steps were positioned less correctly than the extreme steps (Fig. 3 shows the typical U-shaped curve). Moreover, the control subjects performed better than the patients. The frontal patients differed in sequencing the central steps compared with the extreme steps, while

Fig. 2. Experiment 1. Percentage correct scores of frontal lobe patients and control subjects on action sequencing tasks across different conditions.

Fig. 3. Experiment 1. Percentage of action steps correctly positioned in action sequencing tasks. Only frontal lobe patients showed a U-shaped curve: central action steps were less correctly positioned compared with the extreme steps within the action sequence.

the control subjects did not. Post-hoc analyses, using paired sample t-tests for each comparison with Bonferroni corrections for the number of comparisons, revealed that the central steps were positioned significantly less correctly compared with the extreme steps: (first step versus central step: t39 ⫽ –4.96, P ⬍ 0.0001; first step versus last step: t39 ⫽ –1.41, P ⬎ 0.17; central step versus last step: t39 ⫽ –5.62, P ⬍ 0.0001).

Discussion Overall, the performance of the patients on the sequencing tasks remained far below that of the controls, revealing a generalized action sequencing deficit. This behaviour characterized the performance of the patients on all action

94 S. Zanini, R. I. Rumiati and T. Shallice

sequencing tasks irrespective of the type of action (schema versus script) and the modality of item presentation (verbal versus pictorial). Therefore, we confirmed previous findings (Sirigu et al., 1995, 1996, 1998) which involved only verbally presented script sequences, and extended them to pictorial presentations and to actions belonging to the schema level (Schank and Abelson, 1977). The effects were greater with script than with schema stimuli, but when sequences of equal length were compared there were no differences. An analysis of the position at which errors occur was carried out. When the patients were required to rearrange action verbal descriptions or action pictures, they reproduced the outer portions of the action sequences relatively better than the central section. This effect was not found in the control subjects. Two possible explanations of our findings could be that they are due to a disruption of the action representations per se or to a failure in understanding the sequencing tasks. These alternative explanations are assessed in sections 2, 3 and 4, respectively.

Fig. 4. Experiment 2. Percentage of schema actions (the same as selected for experiment 1) correctly produced by frontal lobe patients and control subjects.

omission of an obligatory tool, (d) pantomiming, instead of using the object, the patient pantomimes how it should be used, (e) perplexity, delay and hesitation in starting an action or a subcomponent of an action, (f) toying, brief but repeated touching of an object or objects on the table. The presence of any of these error types in the sequence of action steps led to it being treated as an action error.

Results and discussion Section 2. Schema production Experiment 2 In experiment 2 we asked the subjects to produce 10 routine schema actions which were very familiar to Italian people (the same as in experiment 1). We selected multiple-object actions as we assumed that they would be a more sensitive tool, compared with single-object use tasks, to unmask any deficit in action production. Between three and five objects per action were required for each action to be produced. They were on the table in front of the subjects. No distracting objects were given. Patients with motor impairments due to weakness were helped by the experimenter, as some had limb paresis (e.g. the experimenter held the lower part of the coffee machine whilst the patient was screwing the upper part on to it). No time limits were imposed. The patients were videotaped for subsequent scoring. Action productions were scored using an action error taxonomy utilized in a previous study on two apraxic patients (Rumiati et al., 2002). The following errors were scored: kinematic errors: clumsiness, namely trajectory and grasping errors; sequence errors, namely errors occurring at the temporal sequence of the action: (a) action addition, insertion of a meaningful action step that is not necessary to accomplish the goal of the action, (b) action anticipation, anticipation of an action that would normally be performed later in the action sequence, (c) step omission, omission of a step of the multiple action sequence, (d) perseveration, repetition of an action step previously performed in the action sequence; conceptual errors, namely errors that are concerned with the conceptual aspects of objects and actions: (a) misuse of the object, the object is used in an inappropriate manner, (b) mislocation of the action, the action is appropriate to the object but is performed in the wrong place, (c) tool omission,

As a group, the performance of the patients did not differ from that of the normal control group (patients’ mean % correct actions ⫽ 93.3, SD ⫽ 10; controls’ mean % correct actions ⫽ 95.5, SD ⫽ 7.3; t ⫽ 0.51, P ⬎ 0.6) (see Fig. 4). No deficits in action production were present. In detail, the patients made three action addition errors, one step omission error, two misuse errors and one tool omission error. The control subjects made two step omission errors and three tool omission errors. These findings are not consistent with the view of a deficit of any aspect of action representation in the frontal lobe group. A representational deficit would lead to an impaired performance in action processing irrespective of the task (action sequencing versus action production). Therefore, the action sequencing deficit requires an alternative interpretation.

Section 3. Schema and script generation Experiments 3a and 3b In experiments 3a and 3b the subjects were asked to produce descriptions (i.e. verbal generation task) of 10 schema actions and 10 script actions, respectively (the same as those used in experiment 1). The instruction was as follows: ‘Please describe to me how the following action is performed, step by step’. This test could only be administered to six of the nine patients, as three had a severe Broca’s aphasia. Five of the patients had normal language production and one (number 2) was still able to produce a sufficient amount of language for task performance, as assessed by the Aachen Aphasia Test (Luzzatti et al., 1996) in the clinical evaluation.

Results We submitted the number of action steps generated by the frontal lobe patients and the control subjects to t-tests in

Action sequencing deficit 95

Fig. 5. Experiment 3a. Absolute number of action steps evoked by frontal lobe patients and control subjects during the verbal description of a schema action (the same as selected for experiment 1).

Fig. 6. Experiment 3b. Absolute number of action steps evoked by frontal lobe patients and control subjects during the verbal description of a script action (the same as selected for experiment 1).

order to evaluate whether the patients produced poorer action descriptions verbally. The patients verbally generated the same number of action steps as the controls. No differences between groups were obtained in the total number of schema actions (F1,10 ⫽ 0.31, P ⬎ 0.5) and in the total number of script actions (F1,10 ⫽ 0.004, P ⬎ 0.9). Occasionally the frontal patients added some irrelevant steps to the core action sequence. For instance, in the schema action ‘preparing orange juice’, one patient added action steps such as ‘purchasing the orange’ or ‘washing hands’ and so on. In order to evaluate this phenomenon, we asked two independent judges, who did not participate in the study, to sort the action steps reported by the patients and the control subjects into three classes: (a) relevant steps (i.e. steps necessary to perform the very core of the action, i.e. those used in experiment 1); (b) irrelevant but appropriate steps (i.e. action steps that can be done but do not belong to the very core of the action); (c) irrelevant and inappropriate steps (i.e. action steps that do not belong to the action at all). We compared the mean number of relevant action steps evoked by the patients and the control subjects using t-tests for schema and script actions. No differences were found for the former action type (F1,10 ⫽ 0.47, P ⬎ 0.6) (Fig. 5) or for the latter (F1,10 ⫽ 0.40, P ⬎ 0.7) (Fig. 6). Only two frontal lobe patients added several (in one case seven and in the other case six) irrelevant but appropriate action steps, while the control subjects did not. No irrelevant and inappropriate action steps were added by anyone. Therefore, the results of

experiments 3a and 3b did not depend on the inclusion of additional steps to the correct action descriptions. In addition, unlike their performance on sequencing tasks, the six patients always produced action descriptions in the correct temporal order, respecting the action sequences. For instance, one patient described the script ‘going to the cinema’ as follows: I purchase the ticket, I walk towards the room, I show the ticket, I enter the room, I sit down. By contrast, the rearrangement of the same script for the action sequencing task carried out previously (experiment 1) was as follows: I enter the room, I walk towards the room, I sit down, I purchase the ticket, I show the ticket. There was a dissociation between the performance of the patients on this test and on the action sequencing test (experiment 1). This effect was not due to the selection of the patients. The mean percentage score of the six patients tested in the experiment on the schema sequencing task (experiments 1a and 1b) (mean of pictorial and verbal modality) was 30.1%, while they produced the correct action sequence in 98.3% of actions, on average, for the schema evocation task. Similarly, they scored 30% on the script sequencing task (experiments 1c and 1d) and maintained the appropriate action sequence in 96.7% of actions, on average, on the script evocation task.

Discussion The patients were able to recall from long-term memory both the content (i.e. the correct action steps without including action steps belonging to other schemas or scripts) and the structure (i.e. the correct temporal sequence) of the actions. As far as the results with schema stimuli were concerned, this was true irrespective of the required output method: motor production in experiment 2 and verbal generation in experiments 3a and 3b. These findings suggest that there is no representation deficit for well-learned actions in this group of frontal lobe patients. However, an alternative interpretation could be that the action sequencing impairment is due to a failure to understand the task. This will be considered in the next section.

Section 4. Basic sequencing tasks Experiment 4 (conditions a, b) We asked the nine frontal lobe patients and the nine control subjects to carry out two other sequencing tasks in order to check for the presence of any deficit in understanding the task. In experiment 4a, 10 sets of five geometric shapes (five triangles, five squares, etc.) of different sizes were used. The subjects were asked to rearrange the shapes from the smallest to the largest. In experiment 4b, 10 sets of five Arabic numbers, ranging from 1 to 99, were used. The subjects were required to sequence the numbers from the lowest to the highest. The shapes and the numbers were presented on individual cards and were placed in a pseudo-random manner on the

96 S. Zanini, R. I. Rumiati and T. Shallice

table in both tasks (all the cards had to be moved to carry out the task). No time limits were imposed. The presence of one or more step errors invalidated the action sequence rearrangement.

Results and discussion Both the patients and the control subjects were completely correct on both tasks. No basic sequencing deficits were found in the frontal lobe patients. This rules out the possibility that the action sequencing impairment found was determined by a failure to understand the task.

Of course, it could be argued that the script syntax representations with which Sirigu et al. were concerned relate to MKUs and not to schema-type actions. However, one would then have to explain why our group of frontal patients, who showed the same general type of problem to theirs on ordering script-type actions, also had identical problems in ordering schema-type actions. There were no differences between action types when they were directly compared with the length effect controlled for. In addition, they showed a similar serial position effect within the action sequences, which was not shown in the normal subjects. Thus, it would appear that a set of processes common to temporal ordering in the two tasks was required.

General discussion Working memory

We investigated action processing in frontal lobe patients. We submitted the patients to four action sequencing tasks related to schema-type actions (Schmidt, 1975) and scripttype actions (Schank and Abelson, 1977), presenting each both verbally and pictorially. We found a generalized impairment in action sequencing. Neither a type of action effect nor a modality of item presentation effect was present. We also submitted the patients to two basic sequencing tasks, shape and number sequencing, as Humphreys and Forde (1998) had reported that their individual frontal lobe patients had some sequencing deficit when letters and numbers were used. Our patients performed these control tasks perfectly. Therefore, no basic deficit of sequencing comprehension could account for the action sequencing impairment.

In their 1996 paper, Sirigu et al. put forward the working memory hypothesis. They suggested that the difficulties that pre-frontal patients experienced on the sequencing tasks arose from a working memory deficit (even if assessment of this cognitive function was not carried out), which produces a difficulty in maintaining the finer script structure. Their view is that pre-frontal lesions, sparing the posterior associative cortex where the gross temporal structures of scripts are stored, lead patients to respect only the rough temporal organization of action. However, if the difficulty arises from a working memory problem, it is unclear why the patients can actually produce verbal descriptions of the sequence of actions.

The action representation deficit

Relationship to other studies

Can the theoretical position tentatively put forward by Sirigu et al. (1995) and more definitively by Sirigu et al. (1996, 1998) [see also Humphreys and Forde (1998)] give an explanation of our findings? This view claims that the frontal lobes play a critical role in storing at least some aspects of the representation of well-learned actions, particularly those relating to their temporal organization. In order to investigate further the action representation deficit hypothesis, we asked the patients to produce verbal descriptions of both types of action (schema and script) description. The frontal patients performed as well as the normal controls on these tasks. Moreover, they respected the appropriate sequence of the subactions. This is difficult to interpret within an action representation deficit hypothesis. In this case, the patients should not have been able to evoke the details of the action steps or to respect action sequences. These representations would also appear to include the finer temporal action sequence representations, contrary to Sirigu et al.’s (1996) claim. In addition, we asked the patients to carry out schema actions. Their performance was comparable with the performance of the control subjects. This study suggests that the action representations were intact given that the correct ordering of objects can hardly be carried out by affordance.

It might seem that there is a discrepancy between our script generation findings and those of Sirigu et al. (1995). In fact, Sirigu et al. only tested a single routine action and there was no significant difference between their patient groups on closure or sequence errors with this single action. Moreover, the finding of analogous effects in the sequencing of schema and script actions is very different from that of Sirigu et al. (1998) who used sequences of script and sentence material. However, their finding of a double dissociation between two types of material only requires that one involves specific sequencing processes different from other sequencing operations (Kinsbourne, 1971; Shallice, 1988). In the Sirigu et al. (1998) case it is most plausible that this is the sentence material where performance would be impaired by a specific agrammatism. Our results on action production differ from those of Humphreys and Forde (1998) who also studied patients who had frontal lobe lesions. They reported two patients, FK and HG, who were impaired on action production, action evocation and action sequencing tasks. This generalized action processing deficit was interpreted as an effect of disordered action schemas. They proposed that, within a schema, long-term knowledge of actions and the associated information about their temporal order are intimately linked.

Action sequencing deficit 97

In fact, their patients had large lesions that extended outside the frontal lobes (notably FK had bilateral temporal lesions) and, indeed, FK was only able to name 41% of the objects used in the action production tasks—whether this was due to a pure naming or an agnosic deficit remains unclear. However, it is clear that our patients differed qualitatively from these, possibly due to differences in lesion sites. Thus, our study cannot shed light on the effects they found. One study which is complementary to the present one in producing a double dissociation is that of Rumiati et al. (2002) who described two ideational apraxic patients (DR and FG) with left parietal lesions who had impaired performance on an action production task but showed a normal profile on an action sequencing task. The occurrence of this double dissociation supports the idea that a sequencing deficit has nothing to do with the nature of the underlying action representations. Rumiati et al. proposed to interpret these findings in the light of Norman and Shallice’s (1986) model of willed actions, recently updated by Cooper and Shallice (2000). The impairment in producing well-learned actions was held to be due to a disruption within the cognitive system devoted to schema storage and action selection, namely the contention scheduling system. If one looks in more detail at the occurrence of errors in action sequencing tasks, a major impairment with respect to action temporal organization in the middle of an action sequence was found. In this, our findings replicate those of Sirigu et al. (1996) who found that frontal lobe patients respected, even if only partially, boundaries (i.e. extreme action steps) more than the finer temporal structure of scripts. It seems reasonable to assume that defining the borders of an action sequence requires less resources because these steps are more easily recognizable. To select the appropriate sequence of central steps seems likely to make extensive demands on executive functions. First, it requires the patients to simulate mentally the whole temporal structure of the action. Second, for each element of the sequence the patients must create a correspondence between the verbal or visual token and the internal abstract representation of the action. We presume that the patients are potentially capable of both of these steps from the normal performance in verbal generation of action descriptions (experiments 3a and 3b). However, this does not mean that the patients necessarily carry out these steps satisfactorily when the task implicitly requires it rather than explicitly demands it of them. The third and potentially most critical stage involves the organization of the operations involved in solving the sequencing problem. This is not a routinized skill for most subjects. If, say, the patients consider any two cards at any one time, then he or she must see whether there is or is not a discrepancy with the ‘real’ ordering. If there is a discrepancy, then the pair must be switched round. Most crucially this process must be repeated not just on the cards as yet unchecked, but also on the cards already switched, as they may conflict in their new positions with cards other than the ones with which they had been compared. To organize the

programme of comparisons, and thus of possible switching operations, appropriately would require a novel strategy for most subjects. A more basic process of looking for conflicts between the actual and the required sequence at any particular point would itself require careful monitoring. A difficulty in this third stage, therefore, could well relate to the general problem frontal lobe patients have in avoiding data-driven errors, as in the capture error experiment of Della Malva et al. (1993) or in the sentence completion errors in the Hayling B sentence completion task (Burgess and Shallice, 1996). Such problems would be exacerbated in the present tasks in that the correct response requires the subject to set up an abstract representation of the action sequence which must compete for the patient’s attention with possible plausible responses directly triggered by inappropriate combinations of cards on the table. The processes required of strategy generation and application for a non-routine situation on the one hand and monitoring on the other, are key supervisory subprocesses of the revised supervisory system model of Shallice and Burgess (1996). In that paper the authors considered the key steps in coping with a novel situation. Strategy selection and application on the one hand, and monitoring and checking of the resulting behaviour on the other, were two main processes for which there was evidence of pre-frontal cortical involvement. However, assessment of whether impairment of this third stage or of the earlier two stages creates the sequencing difficulty will require further experimentation. The results overall show that the script sequence phenomena described in frontal patients by Sirigu et al. (1995) are robust. However, in partial disagreement with the position of Sirigu et al. (1998), the effects were found to generalize across different types of action representation. The findings appear to be unrelated to the concept of the MKU or to the storing of finer temporal information concerned with actions. Instead, they relate better to the general ideas of Sirigu et al. (1995) and/or to a disorder of supervisory system subprocesses.

Acknowledgements We would like to thank all the patients for their kind collaboration. We are also grateful to Federica Bearzotti for helping to collect some of the normative data and in preparing some of the stimuli. The research was assisted by support from ‘Cofinanziamento MURST’ (1998) to TS and RIR.

References Arbib MA. Schemas for the temporal organisation of behaviour. Human Neurobiology 1985; 4: 63–72. Bottini G, Corcoran R, Sterzi R, Paulesu E, Schenone P, Scarpa P et al. The role of the right hemisphere in the interpretation of figurative aspects of language. Brain 1994; 117: 1241–53. Burgess PW, Shallice T. Bizarre responses, rule detection and frontal lobe lesions. Cortex 1996; 32: 241–59.

98 S. Zanini, R. I. Rumiati and T. Shallice Cooper R, Shallice T. Contention scheduling and the control of routine activities. Cognitive Neuropsychology 2000; 17: 297–338. Crozier S, Sirigu A, Lehericy S, van de Moortele PF, Pillon B, Grafman J et al. Distinct prefrontal activations in processing sequence at the sentence and script level: An fMRI study. Neuropsychologia 1999; 37: 1469–76. Della Malva CL, Stuss DT, D’Alton J, Willmer J. Capture errors and sequencing after frontal brain lesions. Neuropsychologia 1993; 31: 363–72. Grafman J. Plans, actions, and mental sets: Managerial knowledge units in the frontal lobes. In: Perecman E, editor. Integrating theory and practise in clinical neuropsychology. Hillsdale, NJ: Erlbaum, 1989. Grafman J. Neuropsychology of the prefrontal cortex. In: Zaidel D, editor. Neuropsychology. Handbook of perception and cognition, Vol. 15. San Diego: Academic Press, 1994. Humphreys GW, Forde EME. Disordered action schema and action disorganization syndrome. Cognitive Neuropsychology 1998; 15: 771–811. Kinsbourne M. Cognitive deficit: experimental analysis. In: McGaugh JL, editor. Psychobiology. New York: Academic Press, 1971. Luria AR. Higher cortical functions in man. London: Tavistock, 1966. Luzzatti C, Willmes K, De Bleser R. Aachener Aphasie Test—versione italiana. Firenze: Organizzazioni Speciali, 1996. Norman DA, 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 Press, 1986. Partiot A, Grafman J, Sadato N, Flitman S, Wild K. Brain activation during script event processing. NeuroReport 1996; 7: 761–6. Rumiati RI, Zanini S, Vorano L, Shallice T. A form of ideational apraxia as a selective deficit of contention scheduling. Cognitive Neuropsychology 2001; 18: 617–42. Schank RC. Dynamic memory. Cambridge: Cambridge University Press, 1982. Schank RC, Abelson R. Scripts, plans, goals, and understanding. Hillsdale, NJ: Erlbaum, 1977. Schmidt RA. A schema theory of discrete motor skill learning. Psychological Review 1975; 82: 225–60. Schwartz MF, Reed ES, Montgomery MW, Palmer C, Mayer NH. The quantitative description of action disorganisation after brain damage: A case study. Cognitive Neuropsychology 1991; 8: 381–414. Schwartz MF, Montgomery MW, Fitzpatrick-DeSalme EJ, Ochipa C, Coslett HB, Mayer NH. Analysis of a disorder of everyday action. Cognitive Neuropsychology 1995; 12: 863–92. Schwartz MF, Montgomery MW, Buxbaum LJ, Lee SS, Carew TG, Coslett HB et al. Naturalistic action impairment in closed head injury. Neuropsychology 1998; 12: 13–28. Shallice T. From neuropsychology to mental structure. Cambridge: Cambridge University Press, 1988.

Shallice T, Burgess PW. The domain of supervisory processes and temporal organization of behaviour. Philosophical Transactions of the Royal Society of London B 1996; 351: 1405–12. Sirigu A, Zalla T, Pillon B, Grafman J, Agid Y, Dubois B. Selective impairments in managerial knowledge following pre-frontal cortex damage. Cortex 1995; 31: 301–16. Sirigu A, Zalla T, Pillon B, Grafman J, Agid Y, Dubois B. Enclosing of sequence and boundaries of scripts following prefrontal lesions. Cortex 1996; 32: 297–310. Sirigu A, Cohen L, Zalla T, Pradat-Diehl P, Van Eeckhout P, Grafman J et al. Distinct frontal regions for processing sentence syntax and story grammar. Cortex 1998; 34: 771–8. Spinnler H, Tognoni G. Standardizzazione e Taratusa Italiana di Test Neuropsicologici. Italian Journal of Neurological Sciences 1987 (6).

Appendix 1 List of 10 schema actions Knotting a tie Preparing pasta Preparing a slice of bread with butter and jam Opening a bottle of wine Washing hands Packing a gift Filling a fountain pen Opening a tin of tuna fish Washing up Shaving oneself

Appendix 2 List of 10 script actions Having a meal at the restaurant Going to the cinema Making a telephone call from a public telephone Going swimming at the swimming pool Going to the train station to catch a train Going to the doctor Purchasing a dress Preparing a dinner Going to the market to shop Going to fill the car with petrol

Action sequencing deficit 99

Action sequencing deficit following frontal lobe lesion S. Zanini, R. I. Rumiati and T. Shallice Abstract Frontal lobe patients carried out temporal sequencing tasks related to actions that differed in terms of their abstractness using both verbal and pictorial presentations. A generalized impairment was found: neither a type of action effect nor a modality of item presentation effect was present. The patients also carried out a corresponding action production task and produced actions quickly and without errors. The frontal lobe patients were also spared in generating verbal descriptions of actions: they were as accurate as normal controls both in terms of the details reported and in maintaining the temporal sequence. It has been argued that the difficulty in processing the temporal dimensions of actions following frontal lobe lesions is due to some form of disruption of the action representation. However, no action representational deficits were present in our frontal lobe patients. Thus, they cannot account for our findings. On the contrary, we suggest that the action sequencing deficit was a consequence of the difficulties patients experienced in rejecting wrong alternatives presented by the stimulus situation.

Journal Neurocase 2002; 8: 88–99

Neurocase Reference Number: O247

Primary diagnosis of interest Action sequencing deficit Key words: frontal lobe; sequencing; action processing

Lesion location d Frontal lobe lesions

Lesion type Vascular, traumatic

Language English