Figure 2 shows the percentage of errors made in tactile discrimination judgments. There was a main effect of congruency on the amount of errors participants made in the tactile discrimination task [F(1, 8) = 20.92, p = 0.002, ηp2 = 0.72] and no main effect of modality [F(1, 8) = 2.08, p = 0.187, ηp2 = 0.21]. The fewest errors were made in the Visuotactile Congruent condition (1.13 ± 1.34%), then the Tactile Congruent condition (2.51 ± 2.81%) and Tactile Incongruent condition (3.71 ± 1.35%). The most errors were made in the Visuotactile Incongruent condition (7.91 ± 4.94%). We further found an interaction effect between congruency and modality [F(1, 8) = 5.63, p = 0.045, ηp2 = 0.41] on the percentage of errors made in the tactile discrimination task. Further Bonferroni-corrected (α = 0.025) paired t-testing revealed that the amount of errors was higher for visuotactile incongruent condition than for visuotactile congruent condition [t(8) = − 3.62, p = 0.007, d = − 1.71] but not for the tactile only conditions [t(8) = − 1.37, p = 0.209, d = − 0.64]. This implies that participants made the most mistakes when the non-target object was flashed while the site of tactile stimulation was on the opposite side of the hand. So to summarise, as expected, the amount of errors was larger in the incongruent conditions (tactile and visuotactile) than in the congruent (tactile and visuotactile) conditions, indicating it was harder for participants to judge the site of tactile stimulation when it did not correspond with the side of the hand the non-target object was on. The effect appears to be mostly driven by the visuotactile condition.
Absolute reaction time
We contrasted space, defined as distance from the start button at which the tactile stimulus was given, with mean reaction times across participants (RTs) in Fig. 3 for all conditions. Over all tripwire locations, average reaction times were lower in congruent (312 ± 110 ms) than in incongruent (375 ± 98 ms) conditions [main effect congruency, F(1, 8) = 100.70, p < 0.001, ηp2 = 0.93] but there was no effect of modality and no interaction. As stated in the analysis subsection above, we calculated the slopes of between space and RTs for all participants for each condition. The average of these slopes can be seen in Fig. 3. From this graph, several preliminary conclusions can be made. Most importantly, there seems to be an effect of tripwire location on RTs in the visuotactile congruent condition, but less or none in the other conditions.
Slopes of the fitted linear functions to the data for each condition of each participant showed a main effect of Congruency [F(1, 8) = 13.94, p = 0.006, ηp2 = 0.64]. Slopes were more pronounced in congruent conditions than in incongruent conditions. Furthermore, we found an interaction effect between modality and congruency for the individual slopes for space versus reaction times [F(1, 8) = 6.52, p = 0.034, ηp2 = 0.45]. Further investigation with Bonferroni-corrected paired t-testing (α = 0.025) showed that the individual slopes for visuotactile trials differed between congruent and incongruent conditions [t(8) = − 5.37, p = 0.001, d = -2.53, with steeper slopes in the visuotactile congruent than in visuotactile incongruent condition), which was not the case in the tactile only trials [t(8) = 0.59, p = 0.568, d = 0.28]. Individual one sample t tests (α = 0.0125) showed that the slopes were different from zero only in the visuotactile congruent condition [t(8) = − 5.16, p = 0.001, d = − 1.72] and not in the other conditions [tactile congruent: t(8) = − 1.01, p = 0.341, d = − 0.34, tactile incongruent: t(8) = − 1.58, p = 0.152, d = − 0.53, visuotactile incongruent: t(8) = − 0.08, p = 0.938, d = − 0.03]. Simply put, RTs go down when the hand nears the obstacle, but only when the obstacle is flashed and the tactile stimulation is on the side of the flash. If the tactile stimulation is on the other side or when there is no visual stimulation accompanying the tactile stimulus, the distance between the hand and the obstacle does not modulate RTs.
Cross-modal congruency effect
The cross-modal congruency effect (CCE) measure is shown in Fig. 4. This difference score, which was computed between the congruent and incongruent iterations of tripwire stimulations at the same depth, has been known to show the amount of interference offered by multisensory interactions. That is, the valid stimulation conditions exhibit a facilitatory effect on response times, while the invalid stimulation conditions exhibit an inhibitory effect on response times (see also Spence et al. 2004a, b). Although the exact contribution of each factor (valid and invalid stimulation) remains unclear in this metric, the CCE measure does allow for easy and quantified cross-modal comparisons. We have plotted the CCE against space (depth of stimulation or tripwire location from starting position in mm) for both visuotactile and tactile stimulation conditions. The graphs show an apparent effect of space on the visuotactile CCE, while an effect of space on the tactile CCE appears absent. We ran a least mean squares regression on the average CCE’s across participants per stimulation modality and the results showed that our model had an adjusted r2 of 0.381 in the visuotactile modality condition and an adjusted r2 of 0.0275 in the tactile condition. We analysed the individual slopes of the relationship between space and the CCE for the visuotactile condition (mean ± SD: 0.39 ± 0.21) and the slopes for the relationship between space and the CCE for the tactile condition (mean ± SD: − 0.09 ± 0.49) with a paired t test, which showed a difference between visuotactile and tactile CCE’s [t(8) = − 2.23, p = 0.029 one-sided, d = − 1.05]. Given these data, we can conclude that in the visuotactile condition crossmodal interaction seems to increase when participants’ hands neared the obstacle.
Kinematic data analysis
For the kinematic data analysis we found no main effects for any measure (Peak velocity, Deviation at passing and Error at passing), except for the movement time measure. For movement time we found a main effect of congruency, F(1, 8) = 12.14, p = 0.008, ηp2 = 0.60. This indicates that incongruent trials were performed slower than congruent trials. We found no main effect of modality, but there was a trend for an interaction between modality and congruency [F(1, 8) = 4.66, p = 0.063, ηp2 = 0.37]. There was a trend for the difference in movement time between congruent and incongruent conditions to be larger in visuotactile trials than in tactile only trials [Paired t test: t(8) = − 2.16, p = 0.063, d = − 1.02].
So, the tactile discrimination task seems to influence total movement time on the hand movements, but not other kinematic aspects. However, this change in movement times does not reflect a trade-off with the tactile discrimination task, as participants were slower in incongruent conditions (similar to the tactile discrimination task), especially in the visuotactile condition.
In a control experiment, we placed non-targets at the same distance but to the left of the midline, where no influence of visuotactile interactions is expected, as there is a very low likelihood of touching the non-targets (i.e. they are not obstacles, see Menger et al. 2013a, b). Ten right-handed participants volunteered, but the data of three participants had to be excluded: one participant did not carry out the tactile discrimination task at all, one participant started all hand movements before the start of the trial (and thus before kinematics was measured), and one participant had to be excluded as the marker disconnected during the experiment. The procedure was identical to that of the first experiment, except that as the non-target was placed to the left of the midline, tactile stimuli on the thumb were now considered to be congruent and those of the index finger incongruent.
We performed the same repeated measures ANOVAs as in the main experiment and found no significant effects of congruency or modality in the tactile discrimination task on percentage of errors [modality: F(1, 6) = 1.93, p = 0.214, np2 = 0.24, congruency: F(1, 6) = 1.34, p = 0.292, ηp2 = 0.18, interaction: F(1, 6) = 0.39, p = 0.556, np2 = 0.06) or slopes (reaction times ~ tripwire location) (modality: F(1, 6) = 1.53, p = 0.263, ηp2 = 0.20, F(1, 6) = 2.05, p = 0.202, ηp2 = 0.26, interaction: F(1, 6) = 0.68, p = 0.440, ηp2 = 0.10] and no difference between difference between visuotactile and tactile CCE’s [Paired t test: t(6) = − 0.28, p = 0.787, d = − 0.15]. Please note that while a power analysis (G*power) suggests that a minimum of eight participants is required for detecting a difference, a sequential analysis of the Bayes factor in a Bayesian paired t test (JASP Team 2020) suggested that increasing the sample size would actually result in increasing evidence that the visuotactile and tactile CCE’s do not differ.
Also, there were no effects on average reaction time over all tripwire locations, nor on the kinematic measures Movement time, Deviation at passing or Error at passing. There was a main effect of congruency on peak velocity [F(1, 6) = 6.19, p = 0.047, ηp2 = 0.51] with slightly higher peak velocities in incongruent (120.7 ± 13.6 cm/s) than congruent (119.2 ± 13.4 cm/s) conditions. This shows a slightly different velocity profile depending on congruency, but note that no effects were found on movement time.
As expected, the results of the control experiment do not mirror those of the main experiment. This suggests that the increase in crossmodal interaction when participants’ hands neared the non-target in the main experiment, seen only in the visuotactile condition, indeed depends on the obstructing nature of the non-target.
Distance between the hand and the non-target
In the current experiment, the action was always performed with the right hand and non-targets were placed at the same distance from the midline in the main and the control experiment. But because of the asymmetrical nature of hand movements, there were differences between the experiments in the actual Euclidian distance between the hand and the non-target at the time of the trigger (see Fig. 5). This figure nevertheless shows a considerable overlap between the hand-non-target distances for the main and control experiment. We, therefore, re-analysed only those trials of the main experiment in which the hand-obstacle distance fell within the range of the distances observed for that particular tripwire location in the control experiment. We reran the same repeated measures ANOVA with modality and congruency as factors and reaction time as dependent variable. Due to a loss of power because of the limited number of trials that could be included, the interaction between congruence and modality was no longer significant: modality F(1, 8) = 0.02, p = 0.892, ηp2 < 0.01, congruency F(1, 8) = 3.66, p = 0.092, ηp2 = 0.31, modality*congruency F(1, 8) = 2.84, p = 0.130, ηp2 = 0.26. However, as we had a clear hypothesis about the to-be-expected differences, we decided to perform planned t tests in which differences between congruent and incongruent conditions were tested separately for the visuotactile and tactile modality. They clearly showed that the congruent–incongruent difference remained significant for the visuotactile modality, but not for the tactile modality: visuotacitle: Paired t test: t(8) = − 3.57, p = 0.007, d = − 1.68; tactile only: Paired t test: t(8) = 0.18, p = 0.858, d = 0.09. We suggest, therefore, that the difference in hand–obstacle distance cannot fully explain the difference in crossmodal congruency effects between the main and control experiments.