In all three patient groups, pusher syndrome was assessed by a trained physiotherapist (D.B.) using the standardized 'Scale for Contraversive Pushing' (SCP) 1516. The SCP assesses (i) symmetry of spontaneous posture while sitting and standing, (ii) the use of the non-paretic arm and/or leg to increase pushing force by abduction and extension of extremities while sitting and standing, and (iii) resistance to passive correction of posture while sitting and standing. Pusher syndrome was diagnosed if all three criteria were present and when the patients showed at least a total score of 1 (max. = 2, sitting plus standing) with respect to their spontaneous posture, at least a score of 1 (max. = 2, sitting plus standing) concerning the use of the non-paretic arm and/or leg to increase pushing force by abduction and extension, and when they showed resistance to passive correction of posture. (In four stroke patients, pushing behaviour while standing could not be quantified with SCP due to a complete inability to reach a standing position at the time of the investigation.) Further details of the SCP assessment are presented in Table 1.

Visual field defects were assessed by standard neurological examination. The degree of paresis of the upper and lower limbs was scored with the usual clinical ordinal scale, where '0' stands for no trace of movement and '5' for normal movement. Spatial neglect was diagnosed when the patients showed the characteristic clinical behavior such as orienting towards the ipsilesional side when addressed from the front or the left and/or ignoring of contralesionally located people or objects. In addition, all patients were further assessed with the "Letter cancellation" task 29, the "Bells test" 30, the "Baking tray task" 31, and a copying task 32. Neglect patients had to fulfill the criterion for spatial neglect in at least two of these four clinical tests 33. Aphasia was assessed conducting a bedside examination that evaluated spontaneous speech, auditory and reading comprehension, picture naming, reading, and oral repetition.

The patients with acute unilateral vestibular loss (neuritis, surgery) were admitted with acute vertigo, vomiting and dizziness. They showed severe problems of keeping balance during walking and upright stance with open and closed eyes. The diagnosis of acute unilateral vestibular loss was further based on the following electronystagmographic findings: horizontal spontaneous nystagmus, ability to suppress the spontaneous nystagmus by fixation, no unilateral vestibulo-ocular response to stimulation by rotation of the chair (120°/s) and to unilateral caloric stimulation with warm (44°C), cold (30°C) and ice (3°C) water. Additionally, the subjective visual vertical (SVV) was determined. The selected patients showed an average tilt of the subjective visual vertical (SVV) of 5.0° (SD 3.9°) to the side of the vestibular loss. MR-imaging was performed to exclude ischemia, tumor or any other lesion of the central nervous system in the patients with vestibular loss.

Subjects sat at the bedside with legs hanging free and feet off the ground. They were stabilized by one experimenter standing on the patients' contralesional side or in case of the non brain damaged controls on the left side. The subjects were instructed that they would be moved sideways and that they should not resist this passive movement. They were further instructed to keep arms on their lap and to look into the direction of the camera such that they could easily be recorded from the front (see below). The experimenter applied a smooth and slow sideways tilting motion of the subjects' trunk to the left and right. Subjects were tilted under two viewing conditions for five cycles each, first with open eyes and then while being blind-folded.

White, circular markers (diameter 3 cm) were attached to the subject's body. One marker was placed on the subjects' forehead center, one at the chin, one at the solar plexus, and one at the navel. The first two markers indicated the orientation of the head, the latter two the orientation of the trunk. Two further markers were used to determine the orientation of the lower leg. In stroke patients the markers were attached to the ipsilateral, non-paretic lowerleg. In the non brain damaged control subjects, they were attached to the right lower leg; in the case of acute vestibular patients to the side ipsilateral of the vestibular loss. Marker positions were recorded with a digital camcorder (Sony DCR-HC14E) in a normal lighted room with a sample rate of 30 Hz. The camera was placed at a distance of 3 meters right in front of the subject at a height of 1.20 meters which corresponded to the subject's upper chest. The whole data collection procedure took no longer than ten minutes and usually was performed at the beginning of a regular physiotherapy treatment session.

The white body markers were tracked frame-by-frame using Winanalyze V1.4 (Mikromak), which resembles an automatized software for tracking visual markers in video files. Marker positions were reset manually by one experimenter (L.J.) whenever the software lost its' track. Final marker positions were analyzed using Matlab 6.5 (The Mathworks). Tilt angles of the head, trunk and leg were determined for each single frame. Positive values indicated a rightward or an ipsiversive tilt of the head, trunk or leg with respect to the earth-vertical; negative values a leftward or contraversive tilt. Subsequently, we determined the relative angles of the two segments by calculating the orientation of both the head and the leg with respect to the trunk orientation. Data were averaged across the tilt range for each tilt direction respectively. For each variable, we performed three-way repeated-measures analyses of variance with patient group as the between subject factor and viewing condition and tilt direction as within subject factors. To further analyze significant interactions, we computed additional one- or two-way analyses of variance. For each ANOVA, post-hoc Scheffé comparisons with an α-level of 0.05 were computed between the groups.

To characterize the passive trunk tilt, we determined the maximum trunk tilt range for each individual and performed a fast fourier transformation (FFT) to reveal the frequency of the spectral peak of trunk tilt across the five tilt cycles for each of the two viewing conditions. To obtain trunk tilt velocity (°/sec), we differentiated trunk orientation across two successive frames. The average velocity curve across the five tilt cycles was smoothed with a moving average filter spanning five frames (167 ms). The velocity curve was segmented into individual cycles of movement direction by each zero-crossing. For each direction segment and under both viewing conditions, we determined the maximum trunk velocity.

Since the passive trunk tilt in the roll plane was applied manually by an experimenter, we analyzed the characteristics of the body tilt across patient groups. No effects of viewing condition were found for the tilt range, tilt frequency, or maximum tilt velocity (cf. Table 2).

Due to their strong resistance against passive body tilt when crossing the midline into the ipsiversive direction, the pusher patients demonstrated a smaller average tilt range into the ipsiversive direction compared to the other three control groups (post-hoc comparisons: all p ≤ 0.07). No differences were found between the viewing conditions or between the three subject groups without pusher syndrome. Concerning average tilt frequency, no differences were evident between the four subject groups or the viewing condition. As the interaction between group and viewing condition was significant, we performed separate post-hoc comparisons for both the open and the closed eyes conditions. We found no significant differences (all p ≥ 0.14). Finally, maximum tilt velocity was statistically comparable between the four groups of subjects (all p ≥ 0.40). The results indicate that except for the smaller average tilt range for the patients with pushing behaviour, the characteristics of the passive tilt movement was comparable across all four subject groups.

As the pusher patients were tilted in a smaller range into the ipsiversive direction compared to the other three control groups, we restricted the analysis of the leg-to-trunk orientation (same for the head-to-trunk orientation) to the maximum body tilt angles which were reached by each single participant under both viewing conditions. Thus, the body tilt angles which were analyzed for all subjects ranged from -18.3° (contraversive) to 4.3° (ipsiversive) of body tilt.

Figure 1 gives an example of the behaviour of a pusher patient. He shows a constant ipsiversive tilt of his non-paretic leg (with respect to trunk orientation) when sitting upright as well as during the whole tilt cycle. Figure 2 illustrates the orientation of the leg as a function of trunk tilt for each individual subject of all four groups over the whole tilt cycle. We found significant differences for the average leg-to-trunk orientation over the range of body tilt between the four groups of subjects (F(3,31) = 4.65, p = 0.008). Post-hoc comparisons revealed significant differences between the pusher patients' and the vestibular patients' average leg-to-trunk orientation (p = 0.018), while the difference between the pusher patients and the two other groups did not reach significance (BD-controls: p = 0.078; NBD-controls: p = 0.083). In general, the pusher patients demonstrated a pathological, ipsilateral deviation of their average leg-to-trunk orientation of about 6.9°. Viewing condition did not influence the average leg-to-trunk orientation (F(1,31) = 0.15, p = 0.70). Also, the interaction between group and viewing condition was not significant (F(1,31) = 0.84, p = 0.48). Tilt direction significantly influenced average leg-to-trunk orientation (F(1,31) = 117.96, p < 0.001) and significantly interacted with the group factor (F(3,31) = 11.30, p < 0.001). The interaction between viewing condition and tilt direction was also significant (F(1,31) = 4.62, p = 0.039) as well as the interaction between all three factors (F(3,31) = 4.62, p = 0.009).

Figure 3a presents the average leg-to-trunk orientation for all four groups during passive body tilt into both directions under both viewing conditions. Post-hoc comparisons revealed significant differences between the pusher patients and the other three patient groups during body tilt in the ipsiversive direction with open eyes (all p ≤ 0.018). With closed eyes during body tilt in the ipsiversive direction, the pusher patients' average leg-to-trunk orientation significantly different just from the vestibular patients and the NBD-controls (both p ≤ 0.031). No further comparisons were significant. Two additional paired samples t-tests confirmed that the pusher patients' as well as the BD-controls average leg-to-trunk orientation during ipsiversive body tilt was not affected by viewing condition (both t ≤ 0.82, both p ≥ 0.44). This indicates that the patients with pushing behaviour showed an abnormal average ipsiversive leg-to-trunk orientation of about 9.0° (SD 6.5) when passively tilted into the ipsiversive direction irrespective of visual feedback during the body tilt.

From Figure 2 it is evident, that the pusher patients demonstrated an ipsiversive shift of the intercept as well as a steeper slope of the linear regression of leg orientation as a function of trunk tilt. As for the small tilt ranges (especially into the ipsiversive/rightward direction) and to further describe the leg orientation depending on body orientation, we computed the linear regression parameters for the absolute orientation of the leg as a function of the absolute trunk orientation for each individual and for each body tilt direction and viewing condition. A fit of the linear regression was accepted, if R² exceeded the value of 0.75. Median R² across all individuals and all conditions was high (0.97; range 0.79 to 0.99). For two pusher patients the linear regression parameters could be estimated only for the contraversive side, as goodness of linear fit was below our criterion of 0.75 during body tilt into the ipsiversive direction. Also, one vestibular patient and one NBD-control subject had to be excluded from further analysis because they did not show a sufficient linear fit during body tilt into a particular direction. Table 3 shows the effect table for the analysis of the linear regression parameters for the leg-to-trunk function.

We found a significant effect of the factor group regarding both the intercept and the slope of the linear regression of the leg-to-trunk function. No further main factor and no interaction became significant. Post-hoc comparisons for the intercept parameter revealed significant differences between the pusher patients and all three groups of subjects without pushing behaviour (all p ≤ 0.029). Post-hoc comparisons for the slope parameter showed significant differences between the pusher patients and both the vestibular patients and the NBD-controls (both p ≤ 0.002) and a tendency for the pusher patients to differ from the BD-controls (p = 0.104) indicating the the pusher patients tended to show an increased slope of the leg-to-trunk function. No further comparisons were significant. Thus, we averaged each of the linear regression parameters across the brain damaged controls, the vestibular patients and the non brain damaged controls and across both viewing conditions. From this we determined the relationship between trunk orientation of the pusher patients and the controls for a given leg orientation. For example, while sitting upright the pusher patients demonstrated a spontaneous orientation of the ipsilateral leg that was shown by the non-pusher subjects when tilted on average about 15° to the ipsiversive side.

Analysis of the spontaneous head-to-trunk posture showed a significant effect of group (F(3,31) = 5.79, p = 0.003), a tendency for an effect of viewing condition (F(1,31) = 3.88, p = 0.058), a significant interaction between group and viewing condition (F(3,31) = 3.40, p = 0.03), a tendency for an effect of tilt direction (F(1,31) = 3.36, p = 0.08), no interaction between group and tilt direction (F(3,31) = 0.75, p = 0.53), a significant interaction between viewing condition and tilt direction (F(1,31) = 11.71, p = 0.002), and no interaction between group, viewing condition and tilt direction (F(3,31) = 1.36, p = 0.28). Figure 3b presents the average head-to-trunk orientation for all four groups during passive body tilt into both directions under both viewing conditions. Post-hoc comparisons revealed a significant difference between the BD-controls and the NBD-controls during body tilt in the contraversive/leftward direction with open eyes (p= 0.049). With closed eyes during body tilt into both directions, the NBD-controls' average head-to-trunk orientation was significantly different for the pusher patients' and the BD-controls' average head-to-trunk orientation (all p ≤ 0.023). Additional paired samples t-tests were performed to investigate whether the pusher patients were influenced by the viewing condition. Only during body tilt into the contraversive direction did the pusher patients adopt an increased contraversive tilt of their head-to-trunk orientation with closed eyes (t(8) = 2.43, p = 0.041). The comparison for body tilt into the ipsiversive direction was not significant (t(8) = 1.47, p = 0.18). The opposite was the case for the NBD-controls, they showed an rightward increase of head-to-trunk orientation during rightward body tilt with closed eyes (t(8) = -2.34, p = 0.044) but not during body tilt into the leftward direction (t(8) = -0.56, p = 0.59).

We computed the linear regression parameters for the absolute orientation of the head as a function of the absolute trunk orientation for each individual and for each body tilt direction and viewing condition. Data from three pusher patients were excluded from this analysis because goodness of linear fit was below our criterion of 0.75 during body tilt into the ipsiversive direction. Table 4 shows the effect table for the analysis of the linear regression parameters for the head-to-trunk function.

We found a significant effect of the factor group regarding the intercept of the linear regression of the head-to-trunk function. No further main factor and no interaction became significant with respect to the intercept. Post-hoc comparisons for the intercept parameter revealed significant differences between the BD-controls and the NBD-controls (p = 0.014). For the slope parameter of the head-to-trunk function, we found no effect of group but of viewing condition. The average slope parameter of the head-to-trunk function over all four groups with open eyes was 0.90 (SD 0.27), while it was 0.96 (SD 0.25) with closed eyes. In order to break up the significant interaction between group factor and body tilt direction, we computed two-way repeated-measures analyses of variance on viewing condition and tilt direction for each subject group separately. The pusher patients showed a significant effect of tilt direction (F(1,5) = 11.06, p = 0.02) because of an increased slope (greater than unity) during body tilt into the ipsiversive direction (cf. Table 4) meaning that the tilt of the head was even larger than the trunk tilt. The vestibular patients tended to show the opposite pattern with a greater slope of the head-to-trunk function during body tilt into the contraversive direction (F(1,5) = 3.85, p = 0.097). Interestingly, the vestibular patients were the only group that showed an effect of viewing condition (F(1,6) = 12.72, p = 0.012) which indicated that they kept their head slightly more stable on the trunk with eyes closed (mean = 0.75, SD 0.17) compared to the eyes open (mean = 0.88, SD 0.16).