Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (14)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Hanna-Pladdy, B.
Right arrow Articles by Foundas, A. L.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Hanna-Pladdy, B.
Right arrow Articles by Foundas, A. L.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Brain, Vol. 124, No. 12, 2513-2527, December 2001
© 2001 Oxford University Press

Cortical and subcortical contributions to ideomotor apraxia

Analysis of task demands and error types

Brenda Hanna-Pladdy1,2, Kenneth M. Heilman1 and Anne L. Foundas3

1 Department of Neurology, University of Florida, Gainesville, Florida, 2 Department of Psychology, Louisiana State University, Baton Rouge 3 Deparment of Psychiatry and Neurology, Tulane University School of Medicine, and Neurology Service, Veterans Affairs Medical Center, New Orleans, Louisiana, USA

Correspondence to: Brenda Hanna-Pladdy, Department of Neurology, University of Florida Health Science Center, 100 South Newell Drive, Room L3-100, Box 100236 UFBI, Gainesville, Florida, 32610-0236, USA pladdybh{at}neurology.ufl.edu


   Abstract
Top
 Abstract
Introduction
 Material and methods
 Results
 Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Ideomotor apraxia (IMA) is often associated with damage of the dominant parietal cortex, but many other lesion sites have been implicated suggesting that the praxis system is mediated by a distributed modular network. Although IMA has been reported with subcortical lesions, the role of subcortical structures in the praxis neural network has not been fully addressed. To ascertain the role of subcortical structures in praxis, we compared praxis performance on a variety of tasks in patients with left hemisphere cortical and subcortical lesions. The cortical patients presented with deficits in the production of transitive and intransitive gestures-to-verbal command and imitation, as well as impaired gesture discrimination. In contrast, the subcortical group demonstrated mild production-execution deficits for transitive pantomimes, but normal imitation and discrimination. Qualitative error analysis of production deficits, revealed that both patient groups produced timing errors and the full range of spatial errors. Whereas the subcortical group made more postural errors than the cortical group, sequencing, unrecognizable and no-response errors were only produced by the cortical group. The different profiles of praxis deficits associated with cortical and subcortical lesions, suggests that these structures may have different roles in praxis.

apraxia; ideomotor; limb; basal ganglia; motor control

ANOVA = analysis of variance; AQ = aphasia quotient; BPT = body part as tool; CD = cortical damage; IMA = ideomotor apraxia; MANOVA = multivariate analysis of variance; SCD = subcortical damage; WAB = Western Aphasia Battery


   Introduction
Top
 Abstract
 Introduction
Material and methods
 Results
 Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Hugo Liepmann described several disorders of skilled purposeful movement, ideomotor, limb-kinetic and ideational apraxias (Liepmann, 1905Go, 1920Go). Patients with ideomotor apraxia (IMA) make spatial and temporal errors that are unrelated to elementary motor deficits (Rothi et al., 1988Go; Poizner et al., 1990Go), whereas patients with limb-kinetic apraxia in the absence of weakness present with elemental motor deficits such as slow-stiff movements, and a loss of independent finger movements (Kleist, 1907Go). Patients with ideational apraxia may be impaired in sequencing a series of acts (DeRenzi and Lucchelli, 1988), or make content errors, also referred to as `conceptual apraxia' by some investigators (Ochipa et al., 1992Go).

IMA has been attributed to damage of either the dominant cerebral cortex or the corticocortical connecting pathways (Liepmann, 1900Go, 1905Go; Geschwind, 1965Go; Heilman and Rothi, 1993Go). The postulate that cortical lesions in the suprasylvian and perirolandic region of the left dominant hemisphere induce IMA has received strong support (Kertesz and Ferro, 1984Go; Alexander et al., 1992Go; Schnider et al., 1997Go; for a review, see Leiguarda and Marsden, 2000Go). The failure, however, to identify one specific cortical locus, that when damaged induces IMA, suggests that praxis is mediated by a distributed modular network. Large strokes involving the middle cerebral artery not only extend outside specific cortical areas, but often extend subcortically to periventricular white matter and basal ganglia structures (Laplane et al., 1977Go). Although apraxia arising from `deep' lesions has been repeatedly reported (von Monakov, 1914; Kleist, 1922Go; Agostoni et al., 1983Go; Kertesz and Ferro, 1984Go; Basso and Della Sala, 1986Go; De Renzi et al., 1986Go; Della Sala et al., 1992Go), studies of basal ganglia functions that are based on vascular lesions are complicated due to the uncertainty that the lesion seen on the scan is solely responsible for the observed behavioural dysfunction (Baron et al., 1992Go; Nadeau and Crosson, 1997Go). With infarction, there can be areas of decreased blood flow which are insufficient to cause cystic infarction but are sufficient to cause ischaemic neuronal damage (Lassen et al., 1983Go; Feeney and Baron, 1986Go; Baron, 1987Go, 1989Go; Pappata et al., 1990Go; Heiss, 1992Go; Crosson, 1997Go).

Although there have been a few studies that have systematically examined apraxia following subcortical lesions (De Renzi et al., 1986Go; Della Sala et al., 1992Go), most reports have consisted of isolated individual patients (for a review, see Pramstaller and Marsden, 1996Go). Infarctions limited to the basal ganglia are not common, and other structures such as the thalamus, insula and surrounding white matter are often involved. A review of 82 cases of `deep apraxia' revealed that IMA was most commonly associated with cortical lesions, which extended to the lenticular nucleus or putamen. In these cases, there was often additional involvement of capsular, periventricular or peristriatal white matter. Lesions confined to the basal ganglia (putamen, caudate and globus pallidus) that were associated with apraxia are rare (Pramstaller and Marsden, 1996Go). In contrast, lesions of the thalamus were found to cause apraxia even if there was no apparent involvement of white matter. The pulvinar nucleus of the thalamus and its connections with both the inferior parietal cortex and lateral prefrontal cortex, cortical regions traditionally involved in praxis, have been implicated in cases of thalamic apraxia (Nadeau et al., 1994Go; Shuren et al., 1994Go). In regard to subcortical white matter, the corticocortical fibre pathways which are important for speech and motor control pass through the peristriatal white matter, and it has been proposed that damage to these fibre bundles by deep lesions may account for the associated IMA (Della Sala et al., 1992Go).

If praxis is mediated by an anatomically distributed modular network, injury to specific portions of this network may produce different signs. Unfortunately, many lesion localization studies of apraxia used unidimensional assessments of the praxis system. The failure to include a range of task demands designed to fractionate various modules of the praxis system (Roy et al., 2000Go), has obscured the identification of subtypes of apraxia based on lesion location. This is especially problematic in the investigations of subcortical apraxia, where investigations studied primarily the imitation of intransitive gestures (Basso et al., 1980Go; Agostoni et al., 1983Go; Basso and Della Sala, 1986Go; De Renzi et al., 1986Go; Della Sala et al., 1992Go). Furthermore, these investigations analysed apraxia quantitatively and utilized rigid cut-off scores, rather than conceptualizing apraxia as on a continuum or systematically documenting the qualitative aspects of gestural performances. Qualitative analysis of error types (Rothi et al., 1988Go) might aid in the differentiation between cortical and subcortical apraxia, and help elucidate the role of subcortical structures in mediating skilled acts.

To investigate the existence of apraxia subtypes using error patterns, we evaluated patients with left hemisphere cortical and subcortical ischaemic infarctions. We used MRI images to determine lesion loci and make volume estimates. Furthermore, we only included subjects with small lesions in the subcortical group (primarily striatocapsular infarctions) and excluded patients with thalamic, periventricular white matter, and insular injury as documented by MRI or CT.


   Material and methods
Top
 Abstract
 Introduction
 Material and methods
Results
 Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Subjects
Subjects were recruited from the Tulane University Hospital and Clinics, the New Orleans Veterans Administration Medical Center, and from the Medical Center of Louisiana. For 1 year, patients with a diagnosis of stroke were screened for inclusion. A total of 19 left hemisphere stroke subjects were selected. The criteria for inclusion was lesion location, not the presence of apraxia. Hospital patients without evidence of neurological or psychiatric disease and volunteers from the community comprised the control group (n = 10). Informed consent was obtained from all subjects and the study was conducted in accordance with the Institutional Review Boards of Louisiana State University, Tulane Medical Center and the Veterans Affairs Medical Center, New Orleans. The stroke group had 10 subjects with left hemisphere cortical damage (CD), and nine subjects with left hemisphere subcortical damage (SCD). Left hemisphere-damaged subjects were all screened within 6 months post-stroke, and examined as early as 1 day post-stroke. Since the rapid re-organization of function is characteristic of subcortical structures (Basso et al., 1987Go; Alexander, 1989Go; Mega and Alexander, 1994Go), we evaluated SCD subjects considerably earlier than CD subjects (MCD = 19.4 weeks, SD = 8.7; MSCD = 8.6 weeks, SD = 2.7).

A Kruskal–Wallis one-way analysis of variance (ANOVA) was performed on the demographic variables, and revealed that the three experimental groups (controls, CD and SCD) were not significantly different with respect to age (MControl = 64.10; MSCD = 63.44; MCD = 60.3), sex or race, but were different with respect to education, H(2,29) = 10.562, P = 0.001. Pair-wise comparisons using the Mann–Whitney U-test revealed that the control group had a higher educational level (M = 13.5) when compared with the CD (M = 9.8, P = 0.009) and SCD (M= 8.6, P = 0.004) groups. The years of education for the CD and SCD groups, however, were not different. Age and education were utilized as covariates in analyses where they significantly adjusted for the dependent variables.

Screening
All stroke subjects underwent a complete neurological examination, with a handedness inventory (Briggs and Nebes, 1975Go), a formalized mental status examination (Mini-Mental Status Examination; Folstein et al., 1975Go), a depression screen (Hamilton Depression Rating Scale; Hamilton, 1960Go), and a test of auditory comprehension (Western Aphasia Battery auditory comprehension subtest; Kertesz and Poole, 1974Go). Subjects with poor auditory comprehension, clinically significant depression, or who were not strongly right hand dominant were excluded from the study (see Table 1Go for means and cut-off scores). Several studies have identified that serial sevens and spelling `world' backwards are not interchangeable tasks on the Mini-Mental Status Examination (Watkins et al., 1989Go; Galasko et al., 1990Go). Consequently, instead of calculating Mini-Mental Status Examination scores on either the response to serial sevens or spelling `world' backwards, whichever yielded the higher score, only serial sevens was administered. The usual cut-off score of 24 for dementia on the Mini-Mental Status Examination was adjusted for this variation in test administration, age and education corrected norms (i.e. for subjects older than 60 years of age and with <9 years of education; Tombaugh et al., 1996Go), and failures on items requiring motor participation of the right hemiparetic arm.


View this table:
[in this window]
[in a new window]
  		Table 1 Mean scores (standard deviation) for screening measures
 
Neuroimaging
CT or MRI were obtained for each stroke subject and used for lesion localization. CT scans were performed on a Picker 1200SX Expert or a Picker PQ2000 with a series of 7 mm slices obtained from the foramen magnum to the vertex, and a 2.5 mm interslice gap. MRI scans were performed on a GE Signa 1.5 T magnetic resonance scanner including a sagittal spin echo T1-weighted data set (5 mm thickness, 0.5 mm spacing), axial and coronal fast spin echo T2-weighted data set (5 mm thickness, 1.5 mm spacing), and an axial fast spin echo proton density weighted data set (5 mm thickness, 1.5 mm spacing).

The lesions were mapped using Damasio and Damasio's standardized templates of axial CT/MRI sections at various angles to the canthomeatal line (Damasio and Damasio, 1989Go). Once the lesions were mapped, the location of the lesion and cytoarchitectonic regions involved were determined (Damasio and Damasio, 1989Go). Lesions were classified as cortical, if they involved primarily the grey matter with some extension to the adjacent white matter, whereas subcortical lesions were primarily limited to subcortical grey matter structures (e.g. basal ganglia) and deep subcortical white matter (e.g. periventricular or internal capsule).

Lesion size
Once the lesions were mapped out on standard templates (Damasio and Damasio, 1989Go), these templates were scanned and stored as digital images. ImageJ [a Java image processing program derived from NIH (National Institutes of Health, Bethesda, Md., USA) Image for the Macintosh computer] was utilized to calculate area statistics of the region of interest and the total area of all slices for each template (Rasband, 1998Go; National Institute for Mental Health, Bethesda, Md., USA). Area measurements were created by selection of the wand tool for the total area, and the freehand tool for the region of interest (i.e. lesion). The mean of three trials of freehand measurements was used to calculate each lesion area.

Volumetric estimates of measured areas obtained by the ImageJ processing program were based on the thickness of each cross-sectional area, and an inter-slice gap of 2.5 mm. The total brain volume and lesion volumes for each subject were determined, and then the lesion volume as a percentage of the brain volume was calculated (Table 2Go) by a method developed by the authors (see Appendix I). The mean percentage of brain injury volume was 0.187 (SD = 0.1077) for the SCD and 4.37 (SD = 4.125) for the CD. The CD had significantly larger lesion sizes than the SCD (t = 3.206, P = 0.011) (see Table 2Go).


View this table:
[in this window]
[in a new window]
  		Table 2 Data and lesion summary for each LHD subject
 
Neuropsychological measures
The oral subtests of the Western Aphasia Battery (WAB), spontaneous speech, auditory comprehension, repetition and naming, were administered and the aphasia quotient (AQ) was calculated. Using these scores, the aphasia subtype was determined using the taxonomy suggested by Kertesz (Kertesz, 1979Go, 1982Go).

All subjects were tested with the gesture-to-verbal command subtest of the Florida Apraxia Battery [experimental edition devised in 1992 by L. J. G. Rothi, A. M. Raymer, C. Ochipa, L. M. Maher, M. L. Greenwald and K. M. Heilman (unpublished)]. This subtest contains 20 transitive (tool-use) (e.g. show me how you use scissors to cut paper) and 10 intransitive communicative gestures (e.g. show me how you would salute). The gesture imitation subtest of the Florida Apraxia Battery was also administered. This subtest required subjects to imitate correctly and incorrectly performed transitive and intransitive communicative gestures which were performed by the experimenter. Gesture recognition tests consisting of gesture discrimination and gesture comprehension subtests were also administered. During the gesture comprehension subtest, the subject was required to identify the correct gesture for using a specific tool. The experimenter performs three different gestures, the correct gesture and two foils representing the use of non-target other tools or an intransitive gesture. The gesture discrimination subtest required the subject to select the correctly produced gesture from a series of three gestures, correct versus two foils displaying temporal or spatial aberrations of the correct production.

Subjects were videotaped while performing gestures-to-verbal command and imitation, with the subject's first response for each trial scored by two trained judges on an expanded continuum of severity scale that is similar to Kaplan's 0–3 quantitative scale (Kaplan, 1968Go): 0 (no response, unrecognizable), 1–2 (severely degraded), 3 (moderately degraded), 4–5 (imperfect), 6 (perfect). Qualitative error typing was performed by the same two judges utilizing the scoring system developed by Rothi (Rothi et al., 1988Go), which contains the following four major error categories with specific sub-categorical errors (see Appendix II for definitions of error types): (i) spatial errors [body-part-as-tool (BPT), movement, amplitude, internal and external configuration], (ii) temporal errors (sequencing, timing, occurrence), (iii) content errors (perseverative, related and unrelated content) or (iv) other errors (unrecognizable or no response). We also included an additional movement error: movement direction-incorrect selection of the direction of the rotation (e.g. counterclockwise versus clockwise), and movements in the direction opposite to standard performance (e.g. towards the body versus away from the body). Each apraxic gesture could potentially exhibit more than one type of error.

Apraxia scoring reliability
Inter-rater reliability of movement productions during praxis assessment was determined by correlating the ratings of the first and second raters with a third independent rater. Mean inter-rater agreement regarding the quantitative accuracy score for the gesture-to-verbal command test revealed a significant Pearson Product Moment Correlation of 0.981 (P = 0.003). Inter-rater correlations for each of the error categories (spatial, content and other) across each of the gestures, ranged from 0.905 to 0.985 and were all significant (P = 0.05). Temporal error ratings between raters were less reliable and revealed a Pearson Product Moment Correlation of 0.730. Additionally, inter-rater correlations for each of the error types, revealed high inter-rater reliability for assignment of BPT errors (0.97), internal configuration errors (0.98), and amplitude errors (0.931), significant at the P = 0.05 level. The reliability of assignment of external configuration errors (0.703) and movement errors (0.534) was somewhat lower.


   Results
Top
 Abstract
 Introduction
 Material and methods
 Results
Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Aphasia
Using the scores from the oral language subtests of the WAB and the overall AQ, patients were classified as aphasic or not, and the aphasic patients were classified in aphasic syndromes (see Table 2Go). An ANOVA with AQ as the dependent variable was significant between groups (controls, CD and SCD), F(2,28) = 13.49, P = 0.0001. Pair-wise comparisons with adjustment for multiple comparisons with Bonferroni revealed that the CD group (M = 72.47, SD = 17.9) had significantly lower AQs than control subjects (M = 97.27, SD = 1.99; P = 0.0001) and the SCD group (M = 86.48, SD = 2.76; P = 0.025). The SCD group's AQ was not significantly lower than controls. The SCD subjects with aphasia had anomic aphasia, but the CD group had a variety of aphasic syndromes (see Table 2Go).

Relationship to apraxia
Pearson's correlation coefficient revealed a positive correlation between AQ scores and the apraxia scores from the gesture-to-verbal command (r = 0.782, P = 0.0001) and gesture imitation (r = 0.821, P = 0.001) tests. The apraxia deficit varied according to the classification of aphasia type such that, the SCD aphasic subjects (all classified as anomic) presented with the least praxic deficit, while the CD subjects' scores on the gesture-to-verbal command and imitation tests varied according to their aphasia classification. Anomic aphasics in the CD group (MCommand = 57.64, SDCommand = 15.48; MImitation = 70.67, SDImitation = 9.3), however, did not present with greater praxic deficits when compared with anomic SCD subjects (MCommand = 57.73, SDCommand = 13.77; MImitation = 68.9, SDImitation = 12.2).

Quantitative praxis scores
Transitivity
A multivariate analysis of variance (MANOVA) was performed on the transitive and intransitive gestures comprising the gesture-to-verbal command test, with group as the independent variable, and the total scores of the transitive (pantomimes of using an object or tool) and intransitive communicative gestures as the dependent variables. The MANOVA was significant for group [F(4,52) = 8.58, P = 0.0001], and tests of between-subjects effects were significant for transitive [F(2,26) = 31.02, P = 0.0001] and intransitive gestures [F(2,26) = 5.15, P = 0.013]. Pair-wise comparisons based on estimated marginal means with adjustment for multiple comparisons with Bonferroni revealed that the CD (M = 42.17; P = 0.0001) and SCD patients (M = 48.15; P = 0.0001) when compared with control subjects (M = 82.67) were impaired when performing transitive gestures. The CD group (M = 62.5) also had significantly lower scores than the control group (M = 87.5) when performing intransitive gestures (P = 0.013), but the SCD group (M = 73.89) was not different from controls.

Task demands
A repeated measures MANOVA, with task demands (verbal command, imitation, discrimination, comprehension) as the within-subjects repeated variable, and group (controls, CD and SCD) as the between subjects factor revealed a significant effect for task [F(3,24) = 38.025, P = 0.0001], and a within-subjects main effect for difference in performance between tasks [F(3,78) = 49.28, P = 0.0001]. Tests of within-subjects contrasts for tasks revealed that gesture-to-verbal command produced lower scores than gesture imitation [F(1,26) = 27.52, P = 0.0001], and gesture discrimination produced lower scores than gesture comprehension [F(1,26) = 67.54, P = 0.0001). However, gesture imitation did not produce significantly lower scores when compared with the gesture discrimination score. Between-subjects factors were not interpreted in the repeated measures analyses based on the results of Mauchly's Test of Sphericity ({chi}2 = 4.848), which indicated that the dependent measures were not significantly related.

Two separate MANOVAs were performed on the subject's performance on tests of (i) gesture production (gesture-to-verbal command, and gesture to imitation) and (ii) gesture recognition (gesture discrimination and gesture comprehension). See Table 3Go for means and standard deviations of task demands by group.


View this table:
[in this window]
[in a new window]
  		Table 3 Mean scores (standard deviation) of task demands by group
 
Gesture production
A multivariate analysis of covariance was performed on the dependent measures of gesture-to-verbal command and gesture imitation with group (controls, CD and SCD) as the between-subjects factor. After controlling for education, the multivariate analysis was significant for group [F(4,50) = 3.889, P = 0.008]. Tests of between-subjects effects were significant for gesture-to-verbal command [F(2,25) = 10.727, P = 0.0001], and gesture imitation [F(2,25) = 7.658, P = 0.003]. Pair-wise comparisons based on estimated marginal means and with adjustment for multiple comparisons with Bonferroni revealed that both the CD (P = 0.0001) and SCD (P = 0.04) groups had significantly lower praxis scores on the gesture-to-verbal command test when compared with normal subjects. The CD group (P = 0.002), but not the SCD group, also had greater praxis deficits on gesture imitation.

Gesture recognition
A MANOVA was performed on the dependent measures of gesture discrimination and gesture comprehension, with group as the between-subjects variable. The multivariate was significant [F(4,52) = 2.748, P = 0.038] and the univariates were significant for gesture discrimination [F(2,26) = 5.918, P = 0.008] and gesture comprehension [F(2,26) = 3.332, P = 0.05]. Pair-wise analysis adjusted for multiple comparisons with Bonferroni revealed that only the CD group (CD, P = 0.008; SCD, P = 0.062) was significantly impaired in their ability to discriminate correctly performed gestures from incorrectly performed gestures relative to normal subjects. The ability of the CD group to comprehend the content of a gesture when compared with controls, only approached statistical significance (CD, P = 0.059).

Qualitative error types
A MANOVA was conducted on the specific (i) spatial errors (amplitude, movement, movement direction, BPT, internal and external configuration), (ii) temporal errors (time, sequence, occurrence), (iii) content errors (perseveration, related and unrelated content), and (iv) other errors (unrecognizable and no-response) produced during the gesture-to-verbal command task. For each gesture, qualitative errors were weighted according to the quantitative gesture score, so that the qualitative score reflected not only the number of qualitative errors made, but also the degree to which that particular error distorted the production of the gesture.

The MANOVA revealed a significant difference in qualitative errors between groups (controls, CD and SCD) [F(16,38) = 3.206, P = 0.002] and was followed by univariate tests for the specific qualitative errors. Follow-up comparisons were conducted on the specific errors significant at the univariate level to determine whether stroke groups produced significantly more errors than control subjects.

Spatial errors
Univariate F-tests revealed a significant difference for movement errors [F(2,26) = 30.324, P = 0.0001], movement direction errors [F(2, 26) = 4.338, P = 0.0024], external configuration errors [F(2,26) = 12.67, P = 0.0001], BPT errors [F(2,26) = 9.19, P = 0.001], internal configuration errors [F(2,26) = 4.877, P = 0.016] but not for amplitude errors between groups.

Temporal errors
Univariate F-tests revealed a significant difference for timing errors F(2,26) = 6.987, P = 0.004], and sequencing errors [F(2,26) = 3.414, P = 0.048) between groups. Occurrence errors were not significant between groups.

Content errors
The univariate tests for qualitative content errors (related and unrelated content) and perseverative errors were not significant between groups.

Other errors
The univariate tests for other errors [unrecognizable and no response errors yielding quantitative scores of 0; F(2,26) = 4.064, P = 0.029], were significantly different between groups.

Post hoc analysis
The qualitative error types significant at the univariate level (movement, movement direction, external configuration, BPT, internal configuration, timing, sequencing, and other (unrecognizable and no-response) were investigated with Tukey's HSD post hoc analysis (see Table 4Go for summary of results). This analysis revealed that the CD group made significantly more movement (P = 0.0001), external configuration (P = 0.0001), timing (P = 0.014), BPT (P = 0.034), internal configuration (P = 0046), sequencing (P = 0.049) and other errors (unrecognizable and no response errors; P = 0.025) relative to controls. The SCD group made significantly more movement (P = 0.0001), movement direction (P = 0.035), BPT (P = 0.001), external configuration (P = 0.001), timing (P = 0.006), and internal configuration errors (P = 0.023) when compared with controls.


View this table:
[in this window]
[in a new window]
  		Table 4 Mean difference (standard error) and significance for qualitative errors by LHD cortical and subcortical group
 
Predictors of apraxia severity
Multiple regression was employed to analyse the relationship of task demands to apraxia severity, after controlling for lesion size, and interval between stroke and testing. The dependent measure of apraxia severity was based on a combined total apraxia score across all tasks (gesture-to-verbal command, gesture imitation, gesture discrimination and gesture comprehension). Dummy coding was utilized to categorize subjects into five groups based on their deviation from the mean of controls in terms of apraxia severity. Group 1 included subjects with positive standard deviations or those equal to zero, while the remaining groups included subjects with total apraxia scores which fell below the mean of controls (Group 2, 0 > scores >= –2 SD; Group 3, –2 SD < scores >= –3 SD; Group 4, –3 SD > scores >= –4 SD; Group 5, scores < –4 SD below the mean of controls).

Multiple predictors were entered hierarchically into the regression analysis in two blocks and regressed on apraxia severity. After controlling for age, education, sex and date since stroke in weeks (Block 1), the variables of task demands (gesture-to-verbal command, gesture imitation, gesture discrimination and gesture comprehension), aphasia severity and percentage lesion volume (Block 2) were regressed on apraxia severity. A hierarchical regression was performed and revealed that Step 1 of the regression was significant [F(4,24) = 10.84, P = 0.0001]. Three of the four independent variables, stroke interval (sr2 = 0.241), education (sr2 = –0.125) and sex (sr2 = 0.07), but not age, significantly contributed to the prediction of total apraxia severity. The independent variables in combination contributed another 0.208 in shared variability. Altogether, 64.4% (58.4% adjusted) of the variability in apraxia severity was predicted by knowing the values of these variables.

After Step 1, addition of the independent variables in Block 2 to the equation resulted in a significant increment in R2, and Step 2 of the regression was significant [F(10,18) = 44.36, P = 0.0001]. Three of the six independent variables, gesture to imitation (sr2 = –0.034), gesture-to-verbal command (sr2 = –0.014) and lesion volume (sr2 = 0.012) significantly contributed to the prediction of total apraxia severity. Aphasia severity and the task demands of discrimination and recognition did not significantly contribute. The combination of the six independent variables in Block 2 contributed another 0.257 in shared variability, and contributed another 32% of the variance in total apraxia severity. Altogether, 96.1% (93.9% adjusted) of the variability in apraxia severity was predicted by knowing the values of all of these independent variables (see Table 5Go for summary of regression analysis).


View this table:
[in this window]
[in a new window]
  		Table 5 Hierarchical regression on total apraxia score
 

   Discussion
Top
 Abstract
 Introduction
 Material and methods
 Results
 Discussion
Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
In an attempt to investigate the modular organization of the systems that mediate learned skilled movements and determine the degree of functional segregation of motor control between cortical and subcortical systems, we evaluated left hemisphere-damaged stroke patients for both productive and receptive gestural deficits. The comparison of our results with those of prior studies is difficult to perform because results are heavily dependent on methods, and our methods are different from those used in other studies. These differences include (i) the types of neuropathological processes and lesion loci, (ii) time interval between stroke and testing, (iii) lesion size, (iv) the neurobehavioural tests and (v) the scoring criteria. There are also a paucity of studies that systematically investigated the role of subcortical structures in apraxia. In this discussion we will discuss the role of subcortical structures in an anatomically distributed modular network of limb praxis, but first we will discuss the relationship between aphasia and apraxia.


   Relationship to aphasia
Top
 Abstract
 Introduction
 Material and methods
 Results
 Discussion
 Relationship to aphasia
Quantitative scores
 Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Our findings are in direct agreement with the literature, which suggests that the neural networks that mediate language and praxis overlap. All of our subjects with subcortical lesions who had gesture production deficits also had mild anomic aphasia. In addition, both the CD and SCD subjects classified as having an anomic aphasia demonstrated comparable scores on the gesture-to-verbal command and gesture imitation tasks. Unlike the gesture-to-verbal command, imitation does not require verbal comprehension. Thus, the correlation between performance on gesture to imitation and aphasia severity, mitigates the possibility that the relationship between praxis and language is due to an auditory comprehension disorder.

Once other variables such as stroke–test interval and lesion size were controlled, severity of aphasia accounted for an insignificant amount of the variance in the total apraxia score. Furthermore, the selection of subjects with relatively spared auditory comprehension, and the finding that gesture to imitation and gesture discrimination also proved to be sensitive to impairments of the praxis system from cortical injury, provides evidence that auditory comprehension deficits are not the cause of poor performance on the gesture-to-verbal command task. Our findings indicate that, although there is overlap between apraxia and aphasia, individuals are not apraxic because of language deficits (Goodglass and Kaplan, 1963). The report that apraxia may be present in the absence of aphasia and aphasia may be present in the absence of apraxia (De Renzi et al., 1980Go; Kertesz and Ferro, 1984Go; Papagno et al., 1993Go), suggests that while the modules that mediate these activities overlap anatomically, they are functionally independent.

The dissociation between language and praxis is further supported by cases of verbal dissociation apraxia described by Heilman (Heilman, 1973Go) as well as De Renzi and colleagues (De Renzi et al., 1982Go). These patients, unlike patients with IMA who make spatial and temporal errors, responded to verbal commands with irrelevant responses suggesting that they did not understand the command. Also, unlike patients with IMA who fail to substantially improve with pantomime to imitation, these patients performed flawlessly on imitation and with actual objects, suggesting that their visuokinaesthetic engrams were intact. Furthermore, these patients were capable of verbally describing the movements, and selecting the proper movements from a multiple choice format. These findings suggest that the patients with verbal disassociative apraxia could verbally understand commands for movements, but were unable to access–activate the motor representation called for by the verbal command. These cases illustrate that language and praxis are functionally independent but highly interconnected. Impaired language and the dissociation of language from praxis programmes may interfere with the activation and implementation of the desired programme, but does not cause a dissolution of the praxis programme and thus is not associated with spatial and temporal errors.

Our exclusion of subjects with silent subcortical lesions and significant periventricular white matter disease, decrease the likelihood that praxis deficits in the SCD group are predominantly related to cortical hypoperfusion or infarction not visible on CT or MRI scans. Our findings are also not consistent with the heterogenous group of aphasias reviewed by Nadeau and Crosson (Nadeau and Crosson, 1997Go). In addition to their nonthalamic subcortical lesions, these patients had heterogeneous patterns of cortical hypoperfusion. It is possible that in our patients, white matter injury may have been greater than imaged, but the language impairment of our SCD patients (i.e. intact repetition with anomic aphasia) is the same as those reported by Alexander and colleagues (Alexander et al., 1987Go; Mega and Alexander, 1994Go), whose lesions were confined to the putamen or head of the caudate. Larger lesions extending lateral and superior to the putamen produce greater speech and language disturbances than mild word finding difficulty or hesitancy. These additional aphasic disorders are most likely due to damage to the arcuate fasciculus, the anterior superior periventricular white matter pathways from the supplementary motor area to Broca's area, and the deep anterior frontal white matter which might interrupt anterior callosal fibres (Alexander et al., 1987Go). Although speculative in nature, the pattern of language disturbances suggests that the impairment in the praxis system after subcortical infarction is more likely to be due to the direct consequences of injury to the basal ganglia than to damage of white matter pathways. However, the use of PET or SPECT (single photon emission computed tomography) in future studies might allow a more direct test of the white matter hypothesis.


   Quantitative scores
Top
 Abstract
 Introduction
 Material and methods
 Results
 Discussion
 Relationship to aphasia
 Quantitative scores
Qualitative error patterns
 Appendix I
 Appendix II
 Acknowledgements
 References
 
Transitivity factor
When we attempted to learn if there were differences between our two groups' ability to perform transitive (those that ordinarily involve tool use) and intransitive gestures (those used primarily to communicate), we found that both the CD and SCD groups were impaired for transitive gestures, but only the CD group was impaired when performing intransitive gestures. Consistent with former studies, our findings reveal that transitive gestures are more sensitive to impairment (Goodglass and Kaplan, 1963), perhaps because of their greater complexity (Haaland and Flaherty, 1984Go). The performance of transitive pantomimes requires the integration of motor programmes that control the limb's action in both interpersonal and extrapersonal space. Thus, when performing a transitive act a person has to move the forelimb in respect to the body, and position a part of the forelimb (e.g. hand) so that it is holding/manipulating a tool (e.g. scissors). Transitive gestures, which include the manipulation of objects in space (Mountcastle et al., 1975Go; Taira et al., 1990Go; Sakata et al., 1992Go, 1995Go), may elicit both internal and external configuration errors in both the cortical and subcortical groups (Haaland et al., 1999Go). In contrast, intransitive movements only require the patient to place the hand and arm in a particular relationship to the body (interpersonal space). The dominant parietal cortex has been shown to code the position of meaningful actions in relation to the body (Mountcastle et al., 1975Go; for a review, see Anderson, 1987Go), consistent with the finding that the cortical group was impaired when performing intransitive gestures. Overall, the results support the presence of different categories of `action schemata', with separate categories for object-to-hand movement transformations (see Jeannerod et al., 1995Go; Rizzolatti and Fadiga, 1998Go), and non-tool representational gestures.

The varied effects of transitivity on the gestural performances of our two anatomical groups (CD versus SCD), is supported by prior reports (Dumont et al., 1999Go; Hanna-Pladdy et al., 2001) as well as physiological and functional imaging studies. A recent fMRI investigation revealed that pantomiming tool use recruits activity in additional areas of the cerebral cortex when compared with `tool-independent' hand movements even when the complexity of the task is controlled (Moll et al., 2000Go). Other, brain imaging studies in humans have revealed object-related motor activation guiding the execution of the movement (Perani et al., 1995Go; Martin et al., 1996Go; Grafton et al., 1997Go). The selection of object-oriented responses has been shown to depend on Brodmann area 8, the thalamus, striatum, and deep white matter fascicles of the left hemisphere (Eacott and Gaffan, 1992Go; Petrides et al., 1993Go; Kermadi and Boussaoud, 1995Go; Rushworth et al., 1998Go), while the parietal lobe (Deiber et al., 1991Go, 1996Go; Eacott and Gaffan, 1992Go; Jenkins et al., 1994Go; Toni et al., 1997Go) and area 6 (Rushworth et al., 1997Go; Schluter et al., 1998Go) have been shown to be important in response selection in general (Rushworth et al., 1998Go). The role of the dorsal pathway reaching the premotor cortex in response selection is also supported by a recent PET investigation of meaningful and meaningless hand actions (Grezes et al., 1998Go). Investigations of non-human primates reveals specific neurones selectively responsive to one specific object, even in the absence of visual feedback (i.e. grasping the object in the dark) (Rizzolatti et al., 1988Go), supporting a link between a specific object and actions necessary to manipulate them (Rizzolatti and Fadiga, 1998Go). Our findings are consistent with the existence of preformed motor schemata, which are anatomically linked with cortical and subcortical motor centres.

Task demands
In our study, task demands influenced performance. There was also an interaction between task demands and lesion location (cortical versus subcortical). For both groups, the gesture-to-verbal command task produced significantly lower praxis scores when compared with gesture imitation and both subtests of gesture recognition (discrimination and comprehension). These results are consistent with prior reports and suggest that pantomiming to command is the most sensitive task for apraxia (for reviews, see Heilman and Rothi, 1993Go; Roy et al., 2000Go). Unlike verbal commands, the gesture imitation and recognition/discrimination tests provide the subjects with cues. These cues may aid subjects retrieve the degraded movement representations. Both groups were also more impaired at gesture imitation and discrimination than comprehending gestures. The tasks of verbal command, imitation and discrimination were also sensitive to group differences, while gesture comprehension was not. On the gesture-to-command test, although both the CD and SCD groups were found to have lower means than the control subjects, the CD group was more impaired than the SCD group. When compared with controls, the CD group was impaired when imitating and discriminating gestures, but the subcortical group was not.

These results are difficult to compare with prior studies because different methods were used. Some investigators did not control the means by which gestures were elicited, using mixtures of verbal commands and imitation (Goodglass and Kaplan, 1963; Kertesz and Hooper, 1982Go). Other studies used imitation to circumvent possible comprehension disorders. Scoring methods also vary, with the use of a broader quantitative scale in this investigation and categorical assignment of `apraxic or not' in other studies. These differences can account for some of the variability in results across investigations. Our study suggests that the sole use of imitation (De Renzi et al., 1980Go; Goodglass and Kaplan, 1983Go) may underestimate the degree of praxic impairment. This is a particular problem in the largest apraxia studies of patients with subcortical lesions, where only imitation of intransitive gestures was administered (Basso et al., 1980Go; Agostoni et al., 1983Go; Kertesz and Ferro, 1984Go). Based on this restricted assessment, some of these studies concluded that subcortical structures were not involved in praxis, although a few of these studies were consistent with our results [when tested by verbal command, apraxia can be associated with nonthalamic subcortical lesions (Agostoni et al., 1983Go; for reviews, see Pramstaller and Marsden, 1996Go; Crosson, 1997Go)].

In our investigation, we made an attempt to control for stroke–test interval between groups by testing patients with subcortical lesions earlier after stroke than patients with cortical lesions, consistent with the aphasia literature which suggests that rapid re-organization of function is characteristic of subcortical structures (Basso et al., 1987Go; Alexander, 1989Go; Mega and Alexander, 1994Go). It is unclear, however, whether there is corresponding rapid re-organization of praxis functions after subcortical infarction, and the time of testing within groups was variable. Although time after stroke did not significantly contribute to apraxia severity in our regression equation, it is conceivable that some of the quantitative and/or qualitative differences observed between groups could be the result of time of testing. In the early stages of recovery, there may be brain dysfunction in areas beyond those apparent in anatomical images, while in the late stages the evaluation of symptoms can be complicated by recovery processes. Consequently, further investigation is needed to define the relationship between the rates of recovery of praxic functions and lesion localization.

Several studies suggested that the apraxia associated with subcortical lesions was related to white matter rather than basal ganglia damage (Kertesz and Ferro, 1984Go; Della Sala et al., 1992Go), but apraxia associated with lesions of the dominant basal ganglia have been well documented (Basso and Della Sala, 1986Go; C. A. Kooistra, L. J. G. Rothi, L. Mack and K. M. Heilman, unpublished manuscript, 1991). The one subject in our sample identified as having periventricular white matter involvement demonstrated significant production–execution deficits to verbal command, did not improve on gesture to imitation, and was impaired in gesture discrimination. This subject's pattern of performance was consistent with the cortical rather than the subcortical profile, and emphasizes the importance of considering the involvement of periventricular white matter in cases of subcortical apraxia. Additionally, our findings suggest that lesion size contributes to the severity of apraxia, even after controlling for cortical/subcortical lesion location. This size effect is most likely related to white matter damage, and/or the increased likelihood of damaging multiple areas which are a part of the anatomically distributed modular network. There was, however, considerable variability in lesion size for the CD group, but not the SCD group suggesting that lesion size may have primarily influenced behaviour in the CD group.

With the exception of lesion size, the only predictors, which significantly accounted for the variance in apraxia severity, were the task demands of imitation and verbal command. Conversely, stroke–test interval, which was variable between subjects, did not contribute to apraxia severity. Our findings, along with those of De Renzi and colleagues (De Renzi et al., 1982Go), demonstrate the means of testing apraxia is a critical element in the operational definition of apraxia (Goodglass and Kaplan, 1983Go; Alexander, 1992Go). There is also a growing body of non-human primate and human literature revealing that there are dissociable neural substrates between action execution in the absence of a model verses visual observation and comprehension of actions made by others (Perrett et al., 1989Go, 1990Go; Jeannerod, 1994Go; Rizzolatti et al., 1996Go; Hari et al., 1998Go). Our findings, as well as those of PET and TMS (transcranial magnetic stimulation) studies, support anatomically distinct substrates for action execution and action recognition (Bonda et al., 1994Go; Parsons et al., 1995Go; Rizzolatti et al., 1996Go). This not only supports the visual representation model of praxis, but also supports a model where apraxia can result from (i) the destruction of `visuokinaesthetic' motor representations of learned movement, stored in the posterior association cortex; (ii) from separation of these representations from premotor or motor areas (Heilman, 1979Go; Heilman and Rothi, 1985Go); and (iii) faulty implementation of the innervatory pattern by premotor and motor areas. Our findings extend this model to include subcortical structures, important in the selection of appropriate responses from a general class of possible strategies coded in the innervatory pattern for a particular object. If the visuokinaesthetic motor representations are either damaged or disconnected from premotor or motor areas, then impairment in gesture execution should be evident, regardless of whether elicited by verbal command or a visual model (i.e. gesture imitation) (Heilman et al., 1982Go). With subcortical lesions, the visuokinaesthetic motor representations should be intact and have access to premotor and motor areas. If the primary role of subcortical structures in praxis is the selection of the appropriate components of object-oriented responses, then the presentation of a visual model revealing the appropriate elements of the innervatory pattern should significantly improve performance. This postulate is supported by the SCD group's impairment on gesture-to-verbal command, improved performance when a visual model was provided and, within normal limits, performance on gesture discrimination and comprehension. Although, the quantitative performances between gesture to command and imitation between groups could possibly be due to differences in severity, the postulate that cortical and subcortical structures play different roles in praxis is supported by qualitative error analysis as well as differences in gesture recognition. The SCD group presented with difficulty in selecting the appropriate postural and movement responses (although the general class of motor representations based on knowledge of the tool and object was available), whereas the CD group revealed impairment in the conceptual–semantic aspects of praxis.


   Qualitative error patterns
Top
 Abstract
 Introduction
 Material and methods
 Results
 Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
Appendix I
 Appendix II
 Acknowledgements
 References
 
Qualitative analysis of production errors, revealed that both CD and SCD patients produced timing errors and the full range of spatial errors. Kinematic analysis has demonstrated the tight coupling between spatial and temporal programmes, both of which are disrupted in apraxic patients (Poizner et al., 1990Go; Haaland et al., 1999Go). These findings support the presence of a distributed modular system important for the selection of motor responses and object-oriented responses from spatiotemporal representations, involving left hemisphere cortex, white matter pathways and the striatum (Rushworth et al., 1998Go).

Both CD and SCD groups produced movement, external configuration (orientation) and timing errors. Despite the overall greater praxis impairment associated with damage to the cortex, the SCD patients produced relatively more internal configuration (postural) and BPT errors. Normally, the posture the hand assumes is predicated by the structure of the tool. Cell recording and PET studies are consistent with a role for the striatum and pallidus during the selection of individual object-oriented responses (Kermadi and Boussaoud, 1995Go; Deiber et al., 1996Go; Rushworth et al., 1998Go).

The CD group, however, revealed a failure to retrieve (i) the motor programme (no response and unrecognizable errors), and (ii) the sequence order of complex skilled movements, further supporting the representational hypothesis of apraxia. Movement sequence errors have been attributed to damage of both cortical and subcortical structures. One explanation for inconsistencies in investigations of cortical/subcortical contributions to serial ordering mechanisms may be related to whether the task requires acquisition of a new sequence, or retrieval of a well-learned complex motor sequence. A number of recent investigations have demonstrated that SMA and pre-SMA are prominent in retrieving the memorized information for organizing multiple, discrete movements in a proper sequence to achieve a certain behavioural goal (Halsband et al., 1994Go; Tanji and Shima, 1994Go; Chen et al., 1995Go; Thaler et al., 1995Go; Shima et al., 1996Go; Shima and Tanji, 1998Go), while the basal ganglia have been implicated in the acquisition of new sequential motor programmes (Doyan et al., 1998). Our findings of retrieval difficulties with cortical injury support the proposition of Gabrieli and colleagues (Gabrieli et al., 1993Go), that the long-term representations of learned movements are stored at the cortical level.

Both groups presented with errors in the spatial orientation of movement secondary to improper selection and movement synchronization of joints typical of apraxic impairment (Poizner et al., 1990Go), but only the SCD group produced errors in movement direction. Although numerous studies have proposed different neural circuits that specify movement direction (Rosenbaum, 1980Go; Bonnet et al., 1982Go; Ghez et al., 1990Go; Bock and Arnold, 1992Go; Gordon et al., 1994Go; Fu et al., 1995Go), the neural mechanisms underlying movement direction remain uncertain. The discharge of neurones in motor-related portions of the basal ganglia often reflect the direction of movement. However, lesion and clinical studies do not fully support the concept that basal ganglia contribute substantially to the specification of movement direction (Georgopoulos et al., 1983Go; Crutcher and DeLong, 1984Go; Crutcher and Alexander, 1990Go; Turner and Anderson, 1997Go). Possibly, the relative cortical/subcortical contributions to the control of movement is based on the frequency of changes in movement direction, with greater cortical involvement with more frequent changes (Turner et al., 1998Go). Our results are somewhat inconclusive given that the CD group's production of movement direction errors approached significance, but our tasks did not require frequent movement direction changes, but simply the selection of the proper direction for the trajectory. Consequently, further studies are needed to delineate the relative roles of cortical and subcortical structures in controlling movement direction.

Our results support the postulate that apraxia is characterized by a disruption of spatiotemporal properties of learned skilled movements (Poizner et al., 1990Go; Clark et al., 1994Go; Haaland et al., 1999Go), including the improper translation of spatial plans into angular motions at the joints (Poizner et al., 1995Go). Gestural production to verbal command is impaired after both cortical and subcortical lesions. These spatiotemporal deficits persisted in CD subjects even when language was not used and visual cues or models of the spatiotemporal trajectories were supplied (Poizner et al., 1995Go). CD subjects also had impaired gesture discrimination. Patients with SCD primarily made postural and orientation errors but, with visual cues (imitation) subjects with subcortical lesions improved their postural selections and the orientation of their spatial trajectories. SCD subjects were accurate in their ability to discriminate between correctly and incorrectly performed gestures. These findings suggest that the basal ganglia motor circuit may be involved in the selection of specific individual object-oriented responses of learned skilled movements (Kermadi and Boussaoud, 1995Go; Deiber et al., 1996Go et al., 1997; Rushworth et al., 1998Go) including movement direction (Turner et al., 1998Go). The CD group's gestural impairment after visual cues were supplied as well as the gestural discrimination deficits, support the representational hypothesis of limb ideomotor apraxia (Heilman, 1979Go; Heilman and Rothi, 1985Go).


   Appendix I
Top
 Abstract
 Introduction
 Material and methods
 Results
 Discussion
 Relationship to aphasia
 Quantitative scores
 Qualitative error patterns
 Appendix I
Appendix II
 Acknowledgements
 References
 
Formula for lesion volume
Let N denote the total number of brain images and i the number of images for which a lesion exists. Denote the area of each of these images/lesions by Ai, i = 1, 2, . . ., N. The thickness of each cross-sectional area is denoted by T and the inter-slice gap area between each cross-sectional area as S. The volume, V, is then calculated using the formula:

The S gap is accounted for by using linear interpolation with the average cross-sectional area of the image and the image i + 1[i.e. 1/2(Ai + Ai + 1)]. The lesion volume was computed as a percentage of the total brain volume.


   Appendix II