Exclusion of trials after quality control of the eye movement data

In one neglect patient and one control subject there were technical problems during the presentation of the VS task and the associated recording of the eye movements, which led to the exclusion of all their VS trials from the analysis. Furthermore, single trials were excluded from analysis in all participants, if their number of valid data points (gaze samples on the screen/ presentation duration [ms]) was less than two standard deviations below the group mean. This applied to 5% of the trials in the FV and 4.6% in the VS blocks in the neglect patient group as well as to 6.5% of the trials in FV and 6.1% in VS in the control group.

Outcomes in the FV task

All the control subjects completed the four FV blocks, the patients had less attentional and/or motivational capacities and finished on average 3 (SD = 0.8) blocks. The data of all the relevant outcome measures in the FV and the VS task are presented in Table 2. Furthermore, individual results of the patients are provided in the supplement (Additional file 5: Tables S3–S6).

Table 2 Overview of the resultsCenter of fixation

The CoF in the ORIGINAL condition in the FV task was 6.8° ± 0.8 in the neglect group and 0.3° ± 0.2 in the control group (Fig. 3, Table 2).

Fig. 3

Center of fixation in the free viewing task. The distribution of fixations along the screen’s horizontal axis is depicted as a boxplot diagram, separately for neglect patients and controls as well as for the four different types of modification. The black median band represents the CoF, i.e. 50% of the fixations landed left and 50% right of this position. The boxes span the area from the lower quartile (left end) to the upper quartile (right end) and only 2.5% of all fixations were located outside the whiskers. In neglect patients, the pathological rightward shift of the CoF was significantly reduced under GC-HIGH and GC-LOW as compared to the ORIGINAL version of the picture (*p < 0.05). Further significant between-modification differences are reported in the main text

The ANOVA revealed a significant main effect of MODIFICATION (F (3, 36) = 24.12, p < 0.001), of GROUP (F (1, 38) = 32.51, p < 0.001), and a significant interaction MODIFICATION * GROUP (F (3, 36) = 14.21, p < 0.001).

The patients’ CoF under ORIGINAL was further right than the controls’ (d = 6.6 ± 0.8, p < 0.001). Furthermore, this pathological rightward shift of the patients’ CoF was significantly reduced by both GC modifications (GC-LOW: d = -3 ± 0.4; GC-HIGH: d = -3.2 ± 0.5, both p < 0.00)1, which also differed from STATIC (GC-LOW: d = -2.4 ± 0.3; GC-HIGH: d = -2.6 ± 0.4, both p < 0.001). GC modifications, however, could not completely normalize this bias, as the patients’ CoF under GC-HIGH was still significantly further right than the controls’ CoF under ORIGINAL (d = 3.3 ± 0.9, p = 0.001).

In the control group, the CoF did not differ between the different modifications (always p ≥ 0.846).

First orienting

In the ORIGINAL condition the first saccade was directed leftward in 13.6% ± 3.1 of the FV trials in neglect patients versus 44.6% ± 5.1 in healthy controls (see also Table 2, Fig. 4).

Fig. 4

First orienting in the free viewing task. The first orienting is depicted as the percentage of trials in which the first saccade was directed leftward (i.e. into the neglected hemifield in the patients), separately for the neglect and the control group and the four modification conditions. Under GC-HIGH, neglect patients started their visual exploration more often on the left side of the screen (compared to ORIGINAL, *p < 0.05). The GC modifications also made the controls start more frequently on the left than under ORIGINAL. Further significant between-modification differences are reported in the main text. Error bars show the SEM

The ANOVA for the First orienting in the FV task revealed a significant main effect of MODIFICATION (F (3, 36) = 32.67, p < 0.001) and of GROUP (F (1, 38) = 39.4, p < 0.001). The interaction MODIFICATION * GROUP was not significant (F (3,36) = 2.08, p = 0.12).

In the ORIGINAL condition, patients started their visual exploration significantly less often with a leftward saccade than the controls (d = − 31 ± 6.1, p < 0.001).

Within the neglect group, the percentage of leftward first saccades was higher in GC-HIGH than in all three other modifications (ORIGINAL: d = 21.7 ± 4.8, p < 0.001; STATIC: d = 22.1 ± 4.6, p < 0.001; GC-LOW: d = 12 ± 3.9, p < 0.022). The GC-HIGH modification almost led to a normalization of the patients’ first orienting, as there was no significant difference from the controls’ ORIGINAL condition (d = − 9.3 ± 8.3, p = 0.269).

Within the control group, the first saccade was more often directed leftwards in both GC modifications when compared to the ORIGINAL (GC-LOW: d = 25.8 ± 4.7; GC-HIGH: d = 28.1 ± 4.6, both p < 0.001) and to the STATIC (GC-LOW: d = 20.3 ± 3.7; GC-HIGH: d = 22.5 ± 4.3, both p < 0.001).

Outcomes in the VS task

All the control subjects completed four VS blocks, the patients absolved on average 3.2 (SD = 0.8) blocks.

Center of fixation

In the VS task, the CoF under ORIGINAL was 5.5° ± 0.7 in the neglect group and 1.4° ± 0.2 in the control group (see also Table 2, Fig. 5).

Fig. 5

Center of fixation in de visual search task. The CoF is depicted as the black median band in this box plot diagram on the distribution of all fixations on the screen during the VS task (see Fig. 3 for further explanations). In control subjects, the CoF was significantly shifted to the left under STATIC and GC-HIGH as compared to the ORIGINAL version of the pictures (* p < 0.05). In the Neglect group, this shift was not significant

The ANOVA revealed a significant main effect of MODIFICATION (F (3, 34) = 8.81, p < 0.001) and of GROUP (F (1, 36) = 41.4, p < 0.001), but no significant interaction (F (3, 34) = 0.95, p = 0.427. Under ORIGINAL, the patients’ CoF was further right than the controls’ CoF (d = 3.1 ± 0.7, p ≤ 0.001).

Within the neglect group, there was no significant difference between the different modifications (all p ≥ 0.069).

In the control group, the CoF was significantly shifted further left in STATIC and GC-HIGH as compared to ORIGINAL (STATIC: d = − 1.2 ± 0.3, p = 0.002; GC-HIGH: d = − 1.6 ± 0.5, p = 0.016).

First orienting

In the ORIGINAL condition the first saccade was directed leftward in 7.3% ± 1.8 of the VS trials in neglect patients versus 34.1% ± 3.9 in healthy controls (Table 2, Fig. 6).

Fig. 6

First orienting in the visual search task. Under the two GC-based modifications, neglect patients and control subjects started their visual search significantly more often on the left side of the screen (compared to ORIGINAL, * p < 0.05). The controls were also influenced by the STATIC mask, whereas its impact was not significant in neglect patients. Further significant between-modification differences are reported in the main text. Error bars show the SEM

There was a significant main effect of MODIFICATION (F (3, 34) = 36.58, p < 0.001) and of GROUP (F (1, 36) = 51.87, p < 0.001) on the First orienting (percentage of leftward saccade). The interaction MODIFICATION * GROUP was not significant (F (3,34) = 2.26, p = 0.1).

Patients started less often on the left than the controls under ORIGINAL (d = − 26.8 ± 4.4, p < 0.001).

Within the neglect group, the percentage of first leftward saccades was significantly higher in both GC-modifications than in ORIGINAL (GC-LOW: d = 18.5 ± 4; GC-HIGH: d = 19.2 ± 3.5, both p < 0.001) and in STATIC (GC-LOW: d = 13.6 ± 3.8, p = 0.006; GC-HIGH: d = 14.3 ± 3.3, p = 0.001). Again, the GC-HIGH modification made that the patients’ first orienting approximated the performance of healthy controls under ORIGINAL (d = − 7.6 ± 5.9, p = 0.212).

In the control group, the two GC modifications but also the STATIC modification differed from the ORIGINAL (STATIC: d = 10.3 ± 2.5, p = 0.001; GC-LOW: 24.6 ± 3.6, p < 0.001; GC-HIGH: d = 31.9 ± 3.4, p < 0.001). Again, both GC modifications differed from STATIC (GC-LOW: d = 14.3 ± 3.6, p = 0.002; GC-HIGH: d = 21.6 ± 3.1, p < 0.001).

Omission rate

The neglect patients did not detect 34.7% ± 4.9 of the targets in ORIGINAL, whereas the control participants only missed 1.9% ± 0.7 on average (Table 2, Fig. 7).

Fig. 7

Target omissions in the VS task. The target omission rate [%] is shown separately for the two study groups and the four modification conditions (color-coded lines), in dependence of the four possible target positions (column: OL = outer left, CL = center left, CR = center right, OR = outer right) on the screen. Error bars show the SEM. The color-coded * represents a significant (p < 0.05) difference between the respective modification and the ORIGINAL

The ANOVA revealed a significant main effect of MODIFICATION (F (3, 33) = 5.05, p = 0.005), of COLUMN (F (3, 33) = 31.79, p < 0.001) and of GROUP (F (1, 35) = 51.37, p < 0.001). Furthermore, there was a significant interaction of MODIFICATION * GROUP (F (3,33) = 3.57, p = 0.024) but no significant interaction of COLUMN * MODIFICATION * GROUP (F (9,27) = 1.31, p = 0.279).

For targets located in the outer left under ORIGINAL, patients revealed a greater rate of target omissions than controls (d = 69.6 ± 8.3, p < 0.001). For targets in the outer right under ORIGINAL, the omission rate did not statistically differ between groups (d = 3.2 ± 3.2, p = 0.337).

Within the neglect group, the omission rate in column OL was significantly higher than in column CL (d = 28.4 ± 3.8, p < 0.001), and in column CL higher than in CR (d = 22.2 ± 3.9, p < 0.001) and OR (d = 23.9 ± 5.9, p < 0.001). Regarding the impact of the MODIFICATION, patients had lower omission rates in STATIC than in the other three conditions (all p ≤ 0.014). A closer look at the differences between the modifications per column revealed that these differences were primarily driven by differences in the outer target columns (Fig. 7). Whereas there were no significant differences between the modifications in column CL and CR, in column OL the omission rate was smaller in STATIC than in the other three modifications (all p ≤ 0.022). In column OR, the omission rates were higher under both GC-modifications than under ORIGINAL or STATIC (all p ≤ 0.017).

In the control group the omission rates per column and modification were quite low ranging between 0 and 3.8%. There were no significant differences between target columns or modifications.

Reaction time

The mean RT in ORIGINAL was 6.6 s ± 0.3 for the neglect and 2.5 s ± 0.1 for the control group (Table 2, Fig. 8).

Fig. 8

Reaction times in the VS task. Reaction time [s] until target detection is shown separately for the two study groups and the four modification conditions (color-coded lines) in dependence of the target location on the screen (column OL = outer left, CL = center left, CR = center right, OR = outer right). The color-coded * represents a significant (p < 0.05) difference between the respective modification and the ORIGINAL. Error bars show the SEM

The ANOVA revealed a significant main effect of MODIFICATION (F (3, 33) = 23.86, p < 0.001), of COLUMN (F (3, 33) = 57.14, p < 0.001) and of GROUP (F (1, 35) = 109.84, p < 0.001). The interactions of MODIFICATION * GROUP (F (3,33) = 7.35, p = 0.001) and COLUMN * MODIFICATION * GROUP (F (9,27) = 4.3, p = 0.002) were significant.

For targets in the outer columns (OL or OR) under ORIGINAL, patients yielded higher RTs than the controls (OL: d = 7.6 ± 0.5, p < 0.001; OR: d = 1.2 ± 0.4, p = 0.009).

Regarding the impact of target position, within the neglect group RTs were higher for targets located in OL than in CL (d = 2.4 ± 0.3, p < 0.001) and also higher in CL than in CR (d = 2.4 ± 0.2, p < 0.001) and OR (d = 2.9 ± 0.5, p < 0.001).

Regarding MODIFICATION, the RTs of neglect patients were significantly higher in both GC modifications than in ORIGINAL (GC-LOW: d = 1.3 ± 0.2; GC-HIGH: d = 1 ± 0.2, both p ≤ 0.001) and in STATIC (GC-LOW: d = 1.7 ± 0.2; GC-HIGH: d = 1.4 ± 0.2, both p < 0.001). Figure 8 shows that this difference was predominantly driven by significant differences in CR and OR. For targets located on the right, patients revealed higher RTs under the two GC-modifications than under STATIC (CR and OR: all p ≤ 0.001) and under ORIGINAL (GC-HIGH: CR, d = 1.2 ± 0.4, p = 0.048; OR, d = 2.7 ± 0.5, p < 0.001; GC-LOW: OR, d = 3.3 ± 0.5, p < 0.001).

Within the control group, reactions were slower under GC-HIGH than under STATIC (d = 0.6 ± 0.2, p = 0.015). A closer look at the impact of the target location revealed that this was predominantly driven by differences in column CR, where the RT was significantly higher under GC-HIGH than under STATIC (d = 1.1 ± 0.3, p = 0.027) and also under ORIGINAL (d = 1.5 ± 0.4, p = 0.003).

Single subject analyses of potential responders

We finally looked for potential individual responders among the patients by analyzing on a descriptive single subject basis the impact of the strongest modification (GC-HIGH) on the behavioral performance (omission rate, RT) in the VS task (see Additional file 5: Tables S5 and S6). An individual was defined as a responder if there was a ‘net benefit’ under the modification versus ORIGINAL, i.e. an improvement of the omission rate and/or RT for targets on the left (OL + CL) that was greater than a potential worsening on the right (OR + CR). With respect to target omissions, only four (patient ID 13, 19, 20, 22) out of 19 patients improved under the modification (Additional file 5: Table S5). Regarding RTs, we identified three (patient ID 5, 13, 18) out of 19 patients who yielded such a net benefit under the modification (Additional file 5: Table S6). Taken together, in these descriptive analyses we identified six out of 19 patients who might have benefitted from the GC modification regarding performance in the VS task. When looking at the patient characteristics (age, time since lesion, neglect severity) in this small sample, we could not determine any specific predisposing features, in which these ‘responders’ differed from the rest of the group.

Patients’ report

The patients subjectively rated both the FV task (8.7, SD = 1.6) and the VS task (8.2, SD = 1.9) as highly pleasant. Only 6 of 19 patients reported any notice of the modifications at all, e.g. by stating “image was sometimes less clear” or “seemed as if I’d need glasses”.