Deficient body structural description contributes to apraxic end-position

errors in imitation

a a a,b a
Hormos Salimi Dafsari , Anna Dovern , Gereon R. Fink , Peter H. Weiss

Abstract: Apraxia is a common cognitive deficit after left hemisphere (LH) stroke. It has been suggested that a disturbed representation of the
human body underlies apraxic imitation deficits. Thus, we here tested the hypothesis that a deficient body structural description (BSD), i.e., a
deficient representation of a body part's position (relative to a standard human body), contributes to apraxic end-position errors in imitation,
while controlling for deficits in the semantic representation of the human body (body image, BI) and naming deficits.
A quantitative pointing task to assess putative BSD deficits and an apraxia assessment, including imitation and pantomime tasks, were applied to
27 patients with LH stroke and 19 healthy subjects. While LH stroke patients without apraxia (n=15) did not differ from control subjects in
their pointing performance, patients suffering from imitation apraxia (n=10) showed a differential deficit when pointing to body parts of other
humans compared to object parts. Voxel-based lesion symptom mapping (VLSM) revealed an association of these dif- ferential pointing deficits
(indicating a deficient BSD) with lesions in the angular gyrus of the left inferior parietal cortex.
This first quantitative group study of BSD deficits in LH stroke patients supports the notion that apraxic end- position errors in imitation are – at
least in part – due to a deficient coding of the position of human body parts.

1. Introduction

The impairments caused by apraxia, a cognitive motor deficit that is often observed after left hemisphere (LH) stroke, cannot
be fully ac- counted for by primary sensory and motor deficits, disturbed commu- nication, or lack of motivation (Dovern et
al., 2012). Frequently ob- served apraxic impairments pertain to the imitation of abstract and symbolic gestures, pantomiming
the use of objects and tools, as well as actual object use. The underlying pathophysiology of these apraxic deficits is still
debated.
Goldenberg suggested that apraxic patients are unable “to evoke and represent conceptual knowledge about the human body,
which is necessary for performing the apparently simple task of imitating [meaningless] gestures.” (Goldenberg, 1995). In the
same vein, an in- fluential account proposes an essential role of “body part coding” in apraxia (Goldenberg and Karnath,
2006). These authors stated that “the body parts involved in gestures are categorized” “based on knowledge about distinctive
features and boundaries” of these body parts and that the “gestures are coded as simple spatial relationships between a limited
set of discrete body parts” (see also (Goldenberg, 2009)). Therefore, body part coding is considered “an intermediate step be-
tween perception and reproduction of gestures” (Goldenberg, 1999; Goldenberg and Karnath, 2006). The finding that not only
the imitation but also the matching of gestures is disturbed in LH stroke patients lends further support to the hypothesis that
patients with imitation apraxia may suffer from a deficient conceptual knowledge about the human body (Goldenberg, 1999).
However, studies investigating within the same patient population apraxic deficits and deficits in processing human
bodies/body parts are sparse (Goldenberg, 1995). This is in part because different concepts have been proposed to capture the
neural representation of the human body (Schwoebel and Coslett, 2005): the body schema (BS), the body structural
description (BSD), and the body image (BI).

Body schema (BS) is a „dynamic representation of the relative positions of body parts derived from multiple sensory and
motor in- puts“ (Schwoebel and Coslett, 2005). Head and Holmes first described the BS as a body standard “against which all
subsequent changes of posture are measured before they enter consciousness“ (Head and Holmes, 1911). Note that in many
publications “body schema” is used as an umbrella term for all three terms (body schema, body structural description, and
body image (Le Clec'H et al., 2000)) rather than in its specific meaning, i.e., a dynamic neural representation of the human
body that allows an online updating of the body part's position during limb movement. In contrast, the body structural
description (BSD) is a „topological map of locations derived primarily from visual input that defines body part boundaries
and proximity relationships“ (Schwoebel and Coslett, 2005). Usually, the clinical assessment of the BSD involves asking the
patient to point to a body part on his/her own or another human body (Semenza and Goodglass, 1985). Finally, the body
image (BI, the alternative term used is “body semantics”) is involved in con- sciously evaluating human bodies. BI is regarded
to be „a lex- ical–semantic representation of the body, including body part names, functions, and relations with artefacts“
(Schwoebel and Coslett, 2005). Again, in the psychiatric context, different use of the term “body image” is quite common, e.g.,
body image distortion in anorexia nervosa (Wagner et al., 2003).

Concerning the body part coding hypothesis of imitation apraxia, the BSD is the relevant concept (Goldenberg, 1995). Patients
with a deficient BSD are impaired in localizing a body part (and its bound- aries) on their body or another human's body. Note
that many clinical tests examining imitation deficits in stroke patients (e.g., the Cologne Apraxia Screening, KAS) concentrate
on the end-position of the tested body part (i.e., limb or face) rather than on the (movement) trajectory to this position. This
also applies to the popular Goldenberg imitation tests (imitating hand positions or finger configurations, (Goldenberg, 1995)),
which were used in Goldenberg and Karnath's study on body part coding (Goldenberg and Karnath, 2006). However, in
patients with imitation deficits, dissociations concerning the trajectory and end-po- sition can occur. Some patients show
disturbed trajectories but achieve a proper end-position of the tested effector (e.g., hand), while other patients exhibit a
smooth (movement) trajectory to a wrong end-posi- tion (Hermsdörfer et al., 1996). Thus, BSD deficits may contribute to
apraxic end-position errors during imitation rather than to deficits in (movement) trajectories (Reader et al., 2018).

Nevertheless, BSD deficits could also contribute to object use defi- cits in apraxia. Most likely, a deficient BSD affects both the
panto- miming of object use and the actual use of objects in those cases where the object is used with a given body part (e.g.,
brushing one's teeth, combing one's hair). In contrast, object use that is not directed to the body but instead to another object
(e.g., carving/sawing a piece of wood, cutting a slice of bread) is probably less affected by BSD deficits. Since the contribution
of body part coding and thus BSD to imitation is undisputed (in LH stroke patients), we here concentrate on the re- lationship
between BSD deficits and apraxic end-position errors in imitation. Consequently, we focused on those apraxic patients who
showed end-position errors in imitation in the current apraxia assess- ment (i.e., the Cologne Apraxia Screening, KAS). Note
that the scoring of the KAS imitation subtests is solely based on end-position errors and not also on spatiotemporal errors as in
other apraxia assessments (e.g., the TULIA (Vanbellingen et al., 2009) or the AST (Vanbellingen et al., 2010)). Insights into the
neural correlate of the BSD can be derived from imaging studies in healthy populations and clinical studies in neuro- logical
patients.

Functional imaging studies in healthy subjects implicated different parts of the (left) parietal cortex in the body schema (BS)
and the BSD. In memory-guided reaching movements, a “dynamic representation of the current postural configuration of the
body” (i.e., the BS) was as- sociated with activity in the superior parietal lobe (SPL, (Parkinson et al., 2010)). In contrast, a
task involving the BSD, i.e., pointing to human body parts (compared to pointing to body parts of dogs), acti- vated the
inferior parietal cortex, in particular, the (left) angular gyrus (Felician et al., 2009). Other studies found activations in the left
pos- terior intraparietal sulcus (IPS), when healthy subjects performed tasks involving the BSD (judging the distance between
two body parts (Corradi-Dell'Acqua et al., 2008) or relating the position of a rotated arm to a standard body (Corradi-
Dell'Acqua et al., 2009)).

Concerning the lesions underlying BSD deficits as investigated in studies with neurological patients, the current knowledge is
largely derived from case studies (Buxbaum and Coslett, 2001; Felician et al., 2003). Deficient BSD has oftentimes been
associated with left parietal cortex lesions (Ogden, 1985; Semenza, 1988; Sirigu et al., 1991), a lesion site that is also related
to apraxia (Dovern et al., 2011; Hoeren et al., 2014; Niessen et al., 2014). While there are some behavioural group studies in
neurological patients (Semenza and Goodglass, 1985), a quantitative lesion study examining the (differential) lesion patterns
underlying apraxic end-position errors in imitation and deficits of the BSD is lacking. Therefore, we set out to develop a
quantitative assess- ment of BSD deficits in patients with LH stroke with and without apraxia. Then, these quantitative
behavioural data were subjected to a statistical lesion analysis adopting voxel-based lesion symptom map- ping (VLSM) to
examine the hypothesis that a deficient BSD contributes to apraxic end-position errors in imitation.

2. Methods

2.1. Study participants

Nineteen healthy control subjects (mean age of 59.7 years, range 47–61 years) and 27 patients with first-ever LH stroke (mean
age of 60.3 years, range 26–81 years) were included in the current in- vestigation.
Stroke localisation, including the territory of the medial cerebral artery (MCA) of the LH, was confirmed with the help of
clinical brain imaging by either CT or MRI scans. Imaging was performed within the first days of hospitalisation. Few patients
suffered from a concurrent stroke in adjacent vascular territories of the LH (e.g., posterior (n=2) or anterior (n=1) cerebral
artery strokes). The exclusion criteria were alcohol or drug abuse, uncorrected visual impairment, left-handedness, pre-existing
strokes, and inability to give informed consent. Additionally, patients with clinically relevant depression or dementia
(diagnosed by the treating physicians) were excluded.

Patients and healthy subjects volunteered for the study. The healthy volunteers received a fixed reimbursement of €15 for the
testing (lasting about an hour). The ethics committee of the Medical Faculty, University Hospital Cologne, University of
Cologne had approved the study.

2.2. Clinical and neuropsychological assessment

Apraxia was assessed with the Cologne Apraxia Screening (KAS, (Weiss et al., 2013)). The KAS is a publicly available
screening instru- ment (published by Hogrefe: https://www.testzentrale.de/shop/ koelner-apraxie-screening.html). The KAS
comprises four subtests that are organized in a factorial manner. One factor is task (pantomime versus imitation) and the other
factor is effector (bucco-facial versus limb gestures). This structure results in the following four subtests: 1. pantomime of
object use (i.e., transitive gestures) involving bucco-fa- cial movements; 2. pantomime of object use (i.e., transitive gestures)
involving limb movements; 3. imitation of (intransitive) bucco-facial gestures; 4. imitation of (intransitive) limb gestures. The
stimulus ma- terial consists of photos that either depict the object or a woman showing the to-be-imitated gesture, minimizing
the influence of con- current aphasic deficits (P. H. Weiss et al., 2016, ). All subtests consist of five items each. Moreover, both
imitation subtests contain two meaningless items each (and three meaningful gestures).

In the pantomime subtests, one or two points (depending on the complexity of the pantomime) are awarded for certain
predefined fea- tures of the pantomime. For example, for the item “pantomiming the use of a toothbrush” the following
movement features are scored with one point each: (i) the hand is almost closed to a fist, (ii) the hand is held laterally in front
of the mouth, (iii) the mouth is slightly opened and the teeth are shown, and (iv) circling/pushing movements of the hand.
When a given movement feature is absent, no point is given. For the evaluation of the imitation subtests (i.e., imitation of
bucco-facial and limb gestures), four points are given for the correct imitation on the first trial. Note that an imitated gesture
is considered correct if the patient has adequately reproduced the final end-position of the gesture as depicted on the photos. If
the first imitation trial fails, the stimulus photo is shown for a second time. Two points are given for a successful second trial
or no point for an erroneous second trial. For each of the twenty KAS items, the patient can maximally achieve four points.
Therefore, a maximum score of 20 points can be achieved in each of the four subtests. Thus, the KAS total score can be
maximally 80 points. Based on the normative data of the KAS, a psychometric analysis re- vealed that a patient with a score of
76 or less should be considered apraxic. More information on the KAS can be found in Dovern et al. (2012) and (Kusch et al.,
2018).

As in previous studies (Binder et al., 2017; Dovern et al., 2017), aphasic deficits were assessed by the short version of the
aphasia checklist (ACL-K). The ACL-K contains four subtests: a colour-figure test, a verbal fluency task, a reading task, and a
rating of verbal com- munication (Kalbe et al., 2002). The scores of the ACL-K can amount up to 40 points. The lower the
score, the more severe the aphasic deficits: 14 points or less indicate severe aphasia, scores between 15 and 25 points indicate
moderate aphasia, scores between 26 and 32 indicate mild aphasia, and scores above 33 are within the normal range (i.e., no
aphasia).

(Pre-morbid) handedness was assessed with the help of the “Edinburgh Handedness Inventory” (EHI, (Oldfield, 1971)) for
stroke patients and healthy subjects. The score of the EHI is a laterality quo- tient (LQ), indicating the relative proportion of
actions performed with the right versus the left hand. Therefore, positive EHI-scores (i.e., po- sitive LQs) indicate right-
handedness, while negative EHI-scores (i.e., negative LQs) indicate left-handedness. The higher the positive LQ, the stronger
the right-handedness. The lower the negative LQ, the stronger the left-handedness. Note that patient and control groups were
matched for handedness.
Contralesional limb paresis was evaluated using the Medical Research Council (MRC) scale (O'Brien, 2000).
The MRC-scale is a 5- point-scale indicating the degree of paresis:

➢ Grade 0: No contraction
➢ Grade 1: Flicker or trace of movement
➢ Grade 2: Active movement, with gravity eliminated
➢ Grade 3: Active movement against gravity
➢ Grade 4: Active movement against gravity and resistance ➢ Grade 5: Normal power.

Grades 4-, 4, and 4+ may be used to indicate movement against slight, moderate, and strong resistance, respectively.

Since the experimental procedures required patients to point with their index finger, we asked them to use their left,
ipsilesional hand to avoid putative confounding effects of the contralesional, right-sided paresis. Despite these instructions,
two patients who were not severely paretic (MRC paresis scale of 4+ and 5) decided to use their con- tralesional, right hand
for pointing: one patient (rakl1001m) for both pointing tasks, the other patient (haed1109w) only for the object pointing task.
Patient haed1109w had a perfect score in the object pointing (relative score 1.0) and a nearly perfect score in the body
pointing tasks (relative score 0.98).

The degree of handicap after stroke was assessed using the modified Rankin Scale (mRS; (Rankin, 1957; van Swieten et al.,
1988)). The six grades of this common scale are: 0 = No symptoms at all; 1 = No sig- nificant disability despite symptoms;
able to carry out all usual duties and activities; 2 = Slight disability; unable to carry out all previous activities, but able to
look after own affairs without assistance; 3 = Moderate disability; requiring some help, but able to walk without assistance; 4
= Moderately severe disability; unable to walk without assistance and unable to attend to own bodily needs without
assistance; 5 = Severe disability; bedridden, incontinent, and requiring constant nursing care and attention; 6 = Dead.

Table 1 lists the demographic and neuropsychological data of the healthy controls (n=19), LH stroke patients without apraxia
(n=15), LH stroke patients with apraxia (n=12), and the subgroup of apraxic LH stroke patients with imitation apraxia
(n=10). As a deficient BSD contributes mainly to apraxic end-position errors in imitation, we fo- cused our analyses on those
ten apraxic patients with imitation apraxia in the current apraxia assessment (i.e., Cologne Apraxia Screening, KAS). Detailed
information about the LH stroke patients with apraxia can be found in the supplementary tables.
2.3. Experimental design

Our primary research objective was to investigate specific BSD deficits by assessing pointing to body versus object parts while
con- trolling for naming deficits and deficient body image (BI). We hy- pothesised that LH stroke patients with apraxic end-
position errors in imitation would obtain significantly lower scores in the body part pointing tasks (compared to pointing to
object parts). Such a beha- vioural pattern would suggest that a disturbed BSD is a relevant factor for apraxic end-position
errors in imitation.

After the neuropsychological assessment, subjects were comfortably seated in front of a laptop computer (MacBook Pro), on
which the visual stimuli were presented.

2.3.1. Pointing to body parts

This task aimed to assess the subject's ability to point to parts of the human body. First, we inquired the sections of the human
body, i.e., head, trunk, arms, and legs. Then, we inquired in more detail body parts within these sections (e.g., for the head:
ear, eye, nose, mouth). Fig. 1a depicts a schematic representation of the body part pointing

task. A pre-test aimed at ruling out a body image deficit for the given body part item. Such a body image deficit was
operationalized as the inability to name that body part. Note that impaired body part naming may result from either a
deficient body image or from aphasic deficits. Therefore, participants were asked to name a body part item that was presented
on a computer screen, e.g., a thigh.

If the subjects succeeded in naming that item, they then proceeded to point to the particular item on a picture of a human
body (Fig. 1a). However, if the participant failed to name the shown body part prop- erly, he/she was told the body part's
name. Then, the subject was presented with three related body parts, one of which was the inquired body part, e.g., thigh,
hips, and upper arms. This intermediate step in the testing procedure ensured that patients with aphasic naming pro- blems,
who nevertheless recognized the body part, could still partici- pate in the body part pointing task. If the subject also failed this
task, the given body part item was excluded from the body part pointing task. In other words, if a patient could not name a
body part, she/he was told the name of this body part and was presented with pictures of three related body parts. Then, the
patient was asked to point to the particular body part – selecting it from the three presented body parts. If the patient could
accurately point to the previously named body part indicating that she/he could adequately associate the body part's name
pointing to body part pointing task - although she/he could not name the body part. In contrast, if the patient also failed to
point to the correct body part picture after being told the body part's name, that body part was excluded from the body part
pointing task (see below for the description of the resulting relative performance score).

Note that the inability of LH stroke patients to process a given item that was excluded by the naming pre-test could either
result from a deficient body image or aphasic deficits since both deficits could lead to impaired body part naming. However,
preserved body part naming excludes a relevant deficit of the body image. Thus, the pre-test ensured that a failure in pointing
to human body parts could be considered to reflect a deficient BSD (and not a deficient body image), with lower scores in the
body pointing tasks reflecting a more severe BSD deficit.

For illustration purposes, Fig. 2a depicts a patient who was in- structed to point to the thigh on a picture of a human body
displayed on the monitor. Note that the photo was taken after the patient's response. Therefore, the red and yellow rectangles
used for scoring are already present to illustrate the task and the scoring procedure (see below) within the same figure.

Fig. 2. A. Photograph was taken during the body part pointing tasks (item “thigh”), the patient incorrectly points to the shank. B. Photograph
was taken during the object part pointing tasks (item “lampstand”), the patient correctly points to the stand of the lamp.

2.3.2. Pointing to object parts

This task assessed the participants' ability to point to parts of non- human objects. The procedure was similar to that for the
body part pointing task. Fig. 1b depicts a schematic representation of the object- part pointing task. In each trial, the
participants were asked to name an object that was presented to them on the computer monitor, e.g., a lamp. If the subjects
failed to name the object correctly, they were told the object's name. Then, they were presented with three related objects, one
of which was the inquired object (e.g., lamp, mobile phone, and binoculars), and were asked to select the proper object. This
inter- mediate step in the testing procedure ensured that patients with aphasic naming problems, who nevertheless recognized
the object, could still participate in the object part pointing task. If subjects named or selected the object correctly, they
proceeded to name a part of the given object, e.g., lampstand. If subjects failed in naming the object-part, they were told the
name of the object-part and then instructed to select the given object-part out of an array of three object parts, e.g., lamp stand,
lamp shade, and lamp holder. If subjects succeeded in naming or selecting the object part, they proceeded to point to the
object part on a picture of another object of the given type, e.g., lampstand of another lamp.

Fig. 2b demonstrates that when instructed to point to the lamp- stand, the same patient who failed on the body part pointing
task (Fig. 2a) performed the object pointing task correctly for this test item.

The order of the two pointing tasks was pseudo-randomized, albeit the corresponding naming tasks always preceded the
pointing tasks for either body or object items.

2.3.3. Scoring procedure

A similar scoring system was adopted for both pointing tasks. For

each item, we developed a scoring sheet: A yellow rectangle was closed around the body/object part at which the subject
should point to (e.g., thigh). A red rectangle was closed around the surrounding body parts (e.g., hip and knee). If the subject
correctly pointed to the inquired body/object part within the yellow rectangle, the subject received full points. If the subject
pointed to the body/object part nearest to the inquired part, i.e., within the red rectangle, but outside the yellow rectangle,
the subject received half of the points. If the subject had pointed outside the red rectangle, the subject received no points. The
subjects could score a total of two points for each of the 20 body part items (in total: 2 × 20=40 points) and a total of four
points for each of the ten object part items (in total: 4 × 10=40 points).

Since some patients could not perform the tasks for all items due to their naming deficits (see sections 2.3.1 and 2.3.2), a
relative perfor- mance score was computed reflecting the individual performance in either task (see also Supplementary Table
S3). This relative perfor- mance score was calculated by dividing the score obtained by a given person by the attainable
maximum score for this person in a given task. Thus, if a subject could correctly name or select only 12 items in the body part
naming task, but performed correctly for these 12 items (resulting in 2 points for each of these 12 items), this subject received
a relative performance score of 1 (individual obtained score: 2 × 12=24, individual attainable maximum score: 2 × 12=24,
relative perfor- mance score: 24/24=1). However, if this subject would fail in 3 of the 12 items (resulting in 0 points for these
3 items), then her/his relative performance score would be 0.75 (individual obtained score: 0 × 3 + 2 × 9=18, individual
attainable maximum score: 2 × 12=24, relative performance score: 18/24=0.75).

The subjects were requested to touch the computer screen specifi- cally on the location of their answer (e.g., thigh) and keep
their index finger in that position until the evaluation was carried out by the ex- aminer. The evaluation was performed with
the help of the yellow and red rectangles, as described above. The rectangles appeared on the screen by a button press of the
examiner, while the subjects were asked not to move and to keep their index finger at the location of their an- swer.

2.4. Data analyses

All statistical analyses were performed with the “Statistical Package for the Social Sciences” software (IBM SPSS Version 22.0
for Windows). A repeated-measures ANOVA was calculated to assess the (differential) effects of the two pointing tasks on the
patients' performance. Since the healthy control subjects performed the task “at ceiling” (i.e., without any error in either task),
the data violated the normal distribution re- quirements for an ANOVA. Accordingly, the healthy control subjects’ data were
not included in the ANOVA. As a deficient BSD contributes mainly to apraxic end-position errors in imitation, we focused our
analyses on those ten apraxic patients who showed end-position errors in imitation in the current apraxia assessment (i.e.,
Cologne Apraxia Screening, KAS). Therefore, we computed a 2 × 2 repeated-measures ANOVA for the relative performance
scores with the within-subject factor TASK (body versus object part pointing task) and the between- subject factor GROUP (LH
stroke patients with imitation apraxia [n=10], LH stroke patients without apraxia [n=15]). Results are re- ported at a
significance level of p < 0.05.

Furthermore, correlation analyses between the apraxia test scores (i.e., KAS total score and sub-scores) and the scores of the
two pointing tasks were performed for the same patient sample included in the ANOVA (n=25, i.e., the LH stroke patients
with imitation apraxia [n=10] and the LH stroke patients without apraxia [n=15]).

2.5. Lesion mapping and analyses

Lesion mapping was based on clinical imaging by CT (n=6) or MRI (n=20). For one of the 27 patients (haed1109w) no
imaging suitable for lesion mapping was available since the initial CT scan showed no infarct demarcation despite persistent
neurological/neuropsychological deficits. Note, however, that the pattern of behavioural results was si- milar when the patient
haed1109w was excluded from the analyses. Time from symptom onset to clinical imaging was on average 2.7 days (standard
deviation (SD) 5.6; range 0–26 days).

Using the MRIcron software (http://people.cas.sc.edu/rorden/ mricron/index.html), the investigator (HSD) manually copied
the le- sions to a T1-weighted template brain (ch2.nii). The lesion mapping was double-checked by two further investigators
(AD & PHW); all in- vestigators had to agree on lesion location and extent. At the time of mapping, the two additional
investigators (AD, PHW) were blind re- garding the test performance of a given patient (P.H. Weiss et al., 2016a).
Neuroanatomic localisation of lesion sites was performed using the anatomy toolbox in SPM (Eickhoff et al., 2005).

Voxel-based lesion symptom mapping (VLSM) was employed for statistical lesion analyses (Bates et al., 2003). VLSM was used
to analyse the relationship between lesion site and apraxia severity (oper- ationalized by the KAS total score) as well as
between lesion site and BSD deficits (operationalized as the individual difference score of the body versus object pointing
task). For this difference score, the relative performance scores of the object pointing task were subtracted from the relative
performance scores of the body pointing task in each patient. Only voxels damaged in at least 3 patients (i.e., ≥10% of all
patients) were tested in the VLSM analyses. The statistical threshold was set to p < 0.05 (corrected for False Discovery Rate,
FDR) except when noted otherwise.

3. Results

3.1. Behavioural data

To investigate the hypothesised link between BSD disturbances and apraxic end-position errors in imitation, we evaluated the
relative performance scores in the two pointing tasks between the LH stroke patients with apraxic end-position errors in
imitation (as assessed by the KAS, n=10) and the LH stroke patients without apraxia (i.e., those patients with an
unremarkable KAS score, n=15) using a 2 × 2 re- peated-measures ANOVA.

For the relative performance scores (see Fig. 3), the 2 × 2 repeated- measures ANOVA with the within-subject factor TASK
(body versus object part pointing task) and the between-subject factor GROUP re- vealed a tendency for the main effect of
2
TASK (F(1,23)=3.12, p=0.091, ηp =.119), indicating a trend for higher relative performance scores in the object pointing task
(compared to the body part pointing task). Furthermore, there was a significant main effect of GROUP (F ( 1,23) =5.05, p <
2
0.035, ηp =.180), with lower relative performance scores in the LH stroke patients with imitation apraxia (compared to the
LH stroke patients without apraxia).

2
Importantly, the ANOVA revealed a significant interaction between TASK and GROUP (F ( 1,23)=5.34, p < 0.031, ηp =.188).
This sig- nificant TASK by GROUP interaction resulted from the differentially reduced relative performance score of the LH
stroke patients with imitation apraxia in the body pointing task (0.776, compared to their relative performance score in the
object pointing task: 0.889). In con- trast, LH stroke patients without apraxia performed similarly in both pointing tasks
(relative performance scores: body part pointing task: 0.966, object part pointing task: 0.951).

Note that the LH stroke patients with imitation apraxia who suffered from deficits in pointing to another person's body parts
did also show deficits in pointing to their own body parts when clinically assessed by verbal command (Semenza and
Goodglass, 1985). Due to the lack of a formal assessment of pointing to one's body parts, the current study could not
differentiate BSD deficits related to one's body from those related to other human bodies and whether these contribute to (imi-
tation) apraxia in a differential manner. In their seminal study, Semenza and Goodglass (1985) used different conditions, one
of which (i.e., condition B: The subject points to body parts on a drawing of a human figure on verbal command) resembled
the current body part pointing task. Notably, all examined patient groups of Semenza and Goodglass performed similarly in
condition B as in condition A (see their Table 2 on page 166), in which the patients were asked to point to their body parts on
verbal command. Semenza and Goodglass collapsed conditions A and B for the error analysis (see their Table 1 on page 165).
Consequently, these authors concluded, “that a common factor underlies success in identifying body parts under all
conditions”. Therefore, it can be assumed that this “common factor” (here termed BSD) contributes to apraxic (imitation)
deficits. Furthermore, the cur- rent study adds to the findings of Semenza and Goodglass by including a task that assessed
pointing to object parts allowing for a direct com- parison between body and object part tasks. Thereby, the current data
revealed a specific impairment in pointing to body parts (i.e., BSD deficits) in the current LH stroke patients with apraxic end-
position errors in imitation.

Consistent with the more pronounced severity of aphasic deficits (as assessed by the ACL-K see Table 1) in the apraxic
patients, more items were excluded due to naming deficits, when examining LH stroke pa- tients with imitation apraxia
compared to LH stroke patients without apraxia. Importantly, the amount of excluded items was similar for both pointing
tasks across groups. For the LH stroke patients without apraxia, the number of patients examined with the full set of items
(i.e., 20 items for the body part pointing task and 10 items for the object pointing task) versus a reduced set of items due to
naming deficits was 13 versus 2 for the body part pointing task and 14 versus 1 for the object pointing task. Considering all
apraxic LH stroke patients (n=12, i.e., the 10 patients with apraxic end-position errors in imitation and the two patients with
selective pantomime apraxia), 7 apraxic patients were examined with the full set of items, while 5 apraxic patients were tested
with a reduced item set for both pointing tasks (see Supplementary Table S3).

The single subject data provided in the Supplementary Tables show that BSD deficits are associated with apraxia at the group
level. How- ever, at the single patient level, we also observed dissociations between the performances in the apraxia test and
the pointing tasks (e.g., patient amdi0503m is severely apraxic (and aphasic), but still achieves rea- sonable relative scores of
about 0.6 in both pointing tasks). Concerning aphasia severity, different result patterns were observed. While patients
orro0324m and rakl1001m were severely aphasic (ACL-K-score of 0 and 9, respectively) and performed poorly in the body
pointing task (re- lative performance score in the body pointing task: 0.32 and 0.4, re- spectively), patient lado0805w was also
severely aphasic (ACL-K score of 3), but performed well when pointing to body parts (relative per- formance score in the body
pointing task: 0.93). These result patterns suggest that aphasic deficits observed in our study cannot solely explain the current
BSD deficits.

Correlation analyses (performed for the same patient sample that was included in the ANOVA (n=25), i.e., the LH stroke
patients with imitation apraxia [n=10] and those without apraxia [n=15]) between the apraxia test scores (i.e., KAS scores)
and the scores of the two pointing tasks revealed significant correlations between the relative performance scores of the body
part pointing task and the total KAS score (rho=0.463, p < 0.05) as well as the KAS imitation sub-score (rho=0.450, p <
0.05). Notably, the correlation between the relative performance scores of the body part pointing task and the KAS panto-
mime sub-score revealed a non-significant trend only (rho=0.372, p=0.07). In contrast, there were significant correlations
between the relative performance scores of the object part pointing task and the total KAS score (rho=0.504, p < 0.05) as
well as the KAS pantomime sub-score (rho=0.552, p < 0.05), while the correlation between the relative performance scores
of the object part pointing task and the KAS imitation sub-score revealed a non-significant trend (rho=0.381, p=0.06).
Therefore, the correlation pattern rather supports the notion that BSD deficits (mainly) contribute to apraxic end-position
errors in imitation. However, it should be noted that the correlation results are relatively weak and thus should be interpreted
with caution.
3.2. Lesion analyses

Since the imaging data were available for all but one patient of the current study, Fig. 4A shows the lesion overlay of the 26
LH stroke patients included in the lesion analyses. The highest lesion overlap was observed within the territory of the MCA,
especially in the left middle and superior temporal cortex and the inferior parietal cortex, i.e., an- gular gyrus and
supramarginal gyrus. In contrast, frontal and occipital regions were affected to a lesser degree.

In Fig. 4B, the lesion analysis using voxel-based lesion-symptom mapping (VLSM) for the total score of the Cologne Apraxia
Screening (KAS) is presented. Lesions to the middle and superior temporal gyrus, the operculum, insula, temporal cortex,
supramarginal gyrus, angular gyrus, postcentral gyrus, and the centrally located white matter were significantly associated
with reduced KAS scores and thus the severity of apraxic (pantomime and imitation) deficits (p < 0.05, FDR-cor- rected).
Notably, a very similar lesion pattern for the KAS was observed in a recent study for a group of 44 patients with left
hemisphere (LH) stroke (Binder et al., 2017).

The significant interaction between the two pointing tasks and the factor GROUP (driven by the differentially reduced relative
perfor- mance scores of the apraxic LH stroke patients in the body versus object part pointing tasks) constitutes the main
finding of our study at the behavioural level. Therefore, a further VLSM analysis was performed with the individual difference
scores that were computed by subtracting the relative performance scores in the object pointing task from the relative
performance scores in the body pointing task in each patient. In Fig. 4C, the lesion analysis (VLSM) with these individual
difference scores is presented (p < 0.05, uncorrected). Lesions associated with a worse performance in the body pointing task
(compared to the object pointing task) were mainly found in the angular gyrus of the left in- ferior parietal lobe. Further
smaller lesion sites that showed a similar association were located in the superior and middle temporal cortex.

4. Discussion

To examine the hypothesis that a deficient BSD contributes to apraxic end-position errors in imitation, we developed a test for
quantitatively assessing deficits in pointing to body and object parts that is clinically applicable to LH stroke patients and
minimizes the confounding effects of comorbid aphasia and a deficient body image (i.e., “a lexical–semantic representation of
the body including body part names“ (Schwoebel and Coslett, 2005)). This test revealed that patients with LH stroke and
imitation apraxia are specifically impaired in pointing to body (versus object) parts. Thus, they suffer from a deficient body
structural description (BSD). Within the lesion pattern causing apraxic deficits, the specific deficits in pointing to body parts
(i.e., BSD deficits) were associated with lesions of the angular gyrus of the left inferior parietal lobe. Data support the notion
that apraxic end-position errors in imitation are – at least in part – due to a deficient coding of the position of human body
parts, i.e., a deficient BSD.

To our knowledge, quantitative and clinically applicable assess- ments of the body structural description (BSD, (Semenza,
1988; Semenza and Goodglass, 1985)) or paradigms assessing the body image (BI, (Cash et al., 2004)) are scarce. The current
test procedure ensured that the observed pointing deficits are not solely due to aphasic deficits or a deficient BI since a pre-test
for each item ensured that the subject associated a given body part or object (part) with its name. This al- lowed using a broad
array of body and objects parts.

Regarding the complex definitions of body schema, body image, and body structural description (BSD), we adopted for the
current study the definitions proposed by Schwoebel and Coslett (Schwoebel and Coslett, 2005). These authors consider body
image to be „a lexical–semantic representation of the body, including body part names, functions, and relations with
artefacts“. Accordingly, we operationalized deficits in the body image as deficits in the lexical-semantic representation of body
parts assessed by naming the body parts. Thus, the procedure of ex- cluding body part items from testing (if a given patient
was unable to name a particular body part item AND was unable to correctly point to this body part after the patient was told
the body part's name) mini- mized the impact of aphasic deficits on the current pointing tasks. Importantly, it also excluded a
relevant influence of body image deficits (operationalized as deficits in the lexical–semantic representation of the body parts)
on the performance in the body part pointing task.
Furthermore, the procedures for assessing body and object part pointing were similar so that we could directly compare the
respective relative performance scores for both sub-tests and identify the specific body part pointing deficits of the LH stroke
patients with apraxic end- position errors in imitation. Importantly, the current study, in which aphasic deficits and deficits of
the BI were controlled for, constitutes one of the very few quantitative group studies on BSD deficits (Schwoebel and Coslett,
2005; Semenza and Goodglass, 1985). Most previous studies that carefully delineated BSD from body image (BI) or body
schema (BS) were case reports (Buxbaum and Coslett, 2001; Sirigu et al., 1991).

Fig. 4. A. Lesion overlay of all studied LH stroke patients for whom suitable imaging data was available (n=26). For one of the tested 27
patients no imaging data suitable for lesion mapping was available. Colour shades represent the number of overlapping lesions. The highest
overlap of stroke lesions was observed for the territory of the middle cerebral artery (MCA). B. Lesion analysis adopting voxel-based lesion
mapping (VLSM) of the KAS total scores (n=26). Lesions to the middle and superior temporal gyrus, the operculum, insula, Heschl gyrus,
supramarginal gyrus, angular gyrus, postcentral gyrus, and the centrally located white matter were significantly associated with reduced KAS
scores and thus the severity of apraxic (pantomime and imitation) deficits. Results are displayed at a threshold of p < 0.05, corrected for False
Discovery Rate (FDR). C. Lesion analysis (VLSM) for the differential scores of the body versus object pointing tasks (n=26, un- corrected).
Lesions associated with a worse performance in the body pointing task (compared to the object pointing task) were mainly found in the angular
gyrus of the left inferior parietal lobe. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web
version of this article.)

The relative performance scores in the body part pointing tasks al- lowed us to draw conclusions regarding the severity of the
BSD deficit. Our results showed clear and significant differences in the pointing task performances between the patient groups:
patients with imitation apraxia scored significantly worse in the body pointing tasks than non- apraxic patients, while the
performances of the two groups were similar in the object pointing task. Thus, imitation apraxia was a relevant predictor for
the performance in the pointing task involving human body parts, but not in the pointing task with object parts. Hence, LH
stroke patients with apraxic end-position errors in imitation seem to have a body-selective pointing deficit. As the performance
in the body part pointing task serves as a measure of the severity of a BSD deficit (controlled for general task demands by the
object part pointing task), the current data suggest that LH stroke patients with apraxic end-po- sition errors in imitation have
a specific and more severe BSD deficit than non-apraxic LH stroke patients.
With respect to the lesion analyses with VLSM in the current sample of LH stroke patients (n=26), apraxic deficits in imitation
and panto- mime (as assessed by the KAS) were mainly caused by lesions of the inferior parietal cortex (supramarginal gyrus,
SMG; angular gyrus, AG), the temporal cortex, and the postcentral gyrus. This apraxic lesion pattern is consistent with
previous findings (Dovern et al., 2011; Haaland et al., 2000; Hoeren et al., 2014; Niessen et al., 2014) and very similar to the
lesion pattern reported in a previous study that also used the KAS to test for apraxia (Binder et al., 2017).

BSD deficits (i.e., differential deficits in pointing to body parts versus object parts) were predominantly associated with lesions
of the angular gyrus as part of the inferior parietal lobe. Therefore, within the broader lesions causing apraxia per se, lesions of
the angular gyrus led to specific BSD deficits in the current apraxic LH stroke patients. However, it should be noted that the
VLSM with the difference score body part versus object part pointing did not survive thresholds cor- rected for multiple
comparisons, while the VLSM with the KAS-scores did survive correction for False Discovery Rate (FDR).

The angular gyrus has been associated with different deficits, such as “speech comprehension deficits, finger agnosia, spatial
disorienta- tion, acalculia, agraphia, and dementia” (Seghier, 2013). Note that Goldenberg and Karnath (2006) found that
their patients with selec- tively disturbed imitation of hand postures exhibited a lesion overlap in the inferior parietal lobule
(IPL). These authors argued that especially hand posture imitation draws on body part coding, a notion consistent with the
results of the current quantitative lesion analysis. The current test procedures controlled for the effect of aphasia, which is
reflected in the lesion pattern. While BSD deficits in the current LH stroke patients are associated with inferior parietal cortex
lesions, lesions to the inferior frontal gyrus (Brodmann Area 44) led to combined apraxic and aphasic deficits after LH stroke
(P. H. Weiss et al., 2016). Consistent with the results of our structural lesion analysis, a functional imaging study that
investigated the neural mechanisms underlying the BSD by adopting a task of pointing to human body parts (compared to
pointing to body parts of dogs) also activated the inferior parietal cortex, in particular the (left) angular gyrus (Felician et al.,
2009).

4.1. Limitations

While there has been considerable progress regarding lesion map- ping methods (H.-O. Karnath et al., 2018; Rorden and
Karnath, 2004), some limitations remain or are inherent to the method. One confound is the time since stroke. We used
clinical imaging by CT (n=6) or MRI (n=20) for lesion mapping. As a result, the interval between symptom onset and the
acquisition of the images used was, on average, 2.7 days (standard deviation (SD) 5.6; range 0–26 days). Thus, we used rather
acute images for lesion mapping (H–O. Karnath and Rennig, 2017). Notably, CT was only used for lesion mapping when there
was a de- marcated lesion. Nevertheless, the delineation of the exact lesion extent may have been more difficult than in cases
in which MR images were available. However, the manual tracing technique is still considered to be best suited for exact
lesion delineation (Wilke et al., 2011).

Given the variability in terms of lesion volume, lesion analyses can be performed with co-varying out lesion volume. However,
the dis- advantage of this method is that it only has low statistical power. In fact, “there were no significant voxels for the
logistic regression ana- lyses for sub- and total pantomime and imitation scores when lesion size was added as a covariate” in a
lesion mapping study of 96 LH stroke patients (page 2803 of (Hoeren et al., 2014)). Consistent with this ob- servation, no
voxel was found in the current VSLM analyses when we included lesion volume as a covariate.

In addition to the above-mentioned aspects, there are also issues related to diaschisis (e.g., (Carrera and Tononi, 2014),
(Grefkes and Fink, 2014). Therefore, structural lesion mapping only in part captures the functional consequences of stroke
lesions.

5. Future directions

The current study constitutes one of the very few group studies ex- amining BSD deficits in neurological patients and the first
quantitative lesion study on this topic. Note that our study could not comprehen- sively address all interesting issues
concerning the relationship between BSD and apraxic deficits. Future studies on this topic should explore in more detail
whether a deficient BSD rather contributes to apraxic def- icits in movements directed towards the body (e.g., combing one's
hair) than to apraxic deficits in movements directed away from the body (e.g., cutting a slice of bread). It has been proposed
that especially deficits in imitating meaningless (ML) gestures (compared to mean- ingful (MF) gestures) are related to
deficient “body part coding” (Goldenberg and Karnath, 2006). Accordingly, it is worth to include the meaning of a gesture as a
further dimension in future studies on the relationship between BSD and apraxia to elucidate whether the meaning of a
gesture modulates the effect of a deficient BSD on apraxic imitation deficits (Achilles et al., 2016; Goldenberg and Hagmann,
1997).

6. Conclusion

We have successfully developed a short and clinically applicable test for quantitatively assessing deficits caused by a disturbed
BSD in pa- tients suffering from LH stroke. LH stroke patients with apraxic end- position errors in imitation showed a
differential deficit in pointing to body (versus object) parts. The specific body pointing deficits were mainly associated with
angular gyrus lesions. Thus, our data support the notion that apraxic end-position errors in imitation are – at least in part –
due to a deficient coding of the position of human body parts.