Developmental vision determines the reference frame for the multisensory control of action

Abstract

Both animal and human studies suggest that action goals are defined in external coordinates regardless of their sensory modality. The present study used an auditory-manual task to test whether the default use of such an external reference frame is innately determined or instead acquired during development because of the increasing dominance of vision over manual control. In Experiment I, congenitally blind, late blind, and age-matched sighted adults had to press a left or right response key depending on the bandwidth of pink noise bursts presented from either the left or right loudspeaker. Although the spatial location of the sounds was entirely task-irrelevant, all groups responded more efficiently with uncrossed hands when the sound was presented from the same side as the responding hand (“Simon effect”). This effect reversed with crossed hands only in the congenitally blind: They responded faster with the hand that was located contralateral to the sound source. In Experiment II, the instruction to the participants was changed: They now had to respond with the hand located next to the sound source. In contrast to Experiment I (“Simon-task”), this task required an explicit matching of the sound's location with the position of the responding hand. In Experiment II, the congenitally blind participants showed a significantly larger crossing deficit than both the sighted and late blind adults. This pattern of results implies that developmental vision induces the default use of an external coordinate frame for multisensory action control; this facilitates not only visual but also auditory–manual control.

Environmental events usually stimulate more than one sense. Consequently, spatial information has to be coordinated across different sensory systems to define action goals. A growing body of evidence from animal (1, 2), human (3, 4), and computational modeling (5) studies now suggests that at certain stages of information processing, all visual, auditory, and tactile inputs are mapped into a supramodal frame of reference that predominantly uses an eye-centered or an externally anchored coordinate system^§. It has been suggested that this supramodal spatial representation in turn serves as a “read-out” module for motor systems (1, 5). In line with this hypothesis, many multisensory neurons have been shown to have spatial receptive fields aligned across different sensory modalities (6). When unimodal inputs at near-threshold level are presented bimodally, the response rates of these neurons are sometimes superadditive and performance may often be improved when bimodal events originate from the same location [multisensory facilitation (6, 7)]. Such external coordinate systems seem to be activated by default, even when the use of a modality-specific reference frame would be sufficient to solve a particular task. For example, judgments of the temporal order in which two tactile stimuli are presented to the left and right hand are less accurate when the hands are crossed over the midline (8, 9). This widely reported effect is even observed when different limbs (such as a hand and a foot) are stimulated (10). The crossing effect has been attributed to interference resulting from the misalignment of anatomical (somatotopic) and external frames of reference for coding the location (or side) of tactile stimuli when the limbs are crossed (8, 9).

It has been argued that the default mapping of sensory inputs into external coordinates reflects the dominance of vision in, for example, manual control (1). This view is supported by the fact that auditory spatial maps change with visual maps when animals are raised wearing prismatic spectacles (11). Moreover, there is evidence from both human and animal studies that the capacity of neurons to integrate the information available to different sensory modalities is not innate; rather, it appears to emerge during ontogeny as a consequence of the brain receiving visual inputs (12–14). However, the role of vision for multisensory action control, rather than specifically for sensory perception, has not been investigated thus far.

The present study tested whether developmental vision induces the default use of external coordinates for auditory-manual control. Congenitally blind, late blind, and matched sighted human adults performed an auditory version of the “Simon task” (15). They had to press a left or right response key depending on the bandwidth of pink noise bursts presented from either the left or right loudspeaker. In general, people perform better [lower reaction times (RT) and error rates] when stimuli are assigned to their spatially corresponding response, e.g., a right stimulus to the right hand and a left stimulus to the left hand. This spatial compatibility effect [“Simon effect” (15); see ref. 16 for a comprehensive review] is observed even though the location of the stimulus is of no importance for the task, that is, the participants might, for example, hear tones presented from a loudspeaker on their left or right and have to discriminate the pitch of the tones (i.e., low vs. high) by pressing either the left or the right response key.

The “Simon task” was chosen for the present study because it allows an implicit manipulation of stimulus-response mapping which results in robust (across participants and situations) spatial compatibility effects in healthy humans (17). Spatial compatibility effects as obtained in Simon tasks are considered a “fundamental property of human information processing”(p. 176 of ref. 18) which indicate how stimulus properties (automatically) influence the selection of action (18).

In a second experiment, the same stimuli and experimental settings were used as in Experiment I but the participants were now instructed to respond with the hand located at the position of the sound source. Thus, in this experiment (in contrast to Experiment I), the spatial location of the stimulus was made task-relevant. We hypothesized that the bias to map sensory inputs into external coordinates for action control is induced by developmental vision. Given this assumption, congenitally blind participants, for whom the visual system is absent and consequently cannot trigger the development of such a visually induced bias for activating an external reference frame, should use a stimulus-location/response-hand mapping (i.e., an anatomically anchored reference frame) as a default, whereas sighted controls should, as suggested by earlier findings (15) presumably invoke a stimulus-location/response-location mapping (i.e., an external reference frame) as a default instead. To test these different stimulus-response mappings, both Experiment I and Experiment II were performed with uncrossed and crossed hands. The congenitally blind (but not the sighted) were expected to show a reversed Simon effect in Experiment I when completing the task with their hands held in a crossed posture, and secondly, a larger performance decrement with their hands crossed in Experiment II where the location of the sound source and hand position had to be mapped explicitly. Finally, we hypothesized that during ontogeny vision induces irreversible changes in the functional organization of multisensory brain systems with the result that late blind individuals use similar default reference frames for action control as the sighted.

Results

Experiment I.

When the participants' hands were uncrossed, similar results were obtained in all groups (see Fig. 1 a and c): performance, as indicated by inverse efficiency scores [IE; correspond to corrected RTs (19)], was superior when the target stimulus and response locations spatially corresponded as compared with when they did not. In contrast, when performing the task with crossed hands, the Simon effect (i.e., the spatial compatibility effect) was reversed (with respect to an external frame of reference) in the congenitally blind, but not in either the sighted or late blind groups. Thus, whereas sighted and late blind participants used a stimulus-location/response-location mapping, the congenitally blind invoked a stimulus-location/response-hand mapping (Congenitally Blind vs. Sighted: Visual Status × Hand Posture × Spatial Correspondence interaction: F_1,20 = 19.63, P < 0.001; the same interaction for the comparison of Late Blind vs. Sighted: not significant; see Table 1). In other words, an auditory stimulus presented from a particular loudspeaker (e.g., the left one) primed the hand situated closest to that loudspeaker, irrespective of whether it was the anatomical left or right hand in the sighted and late blind groups. By contrast, the same sound primed the anatomical left hand, regardless of where in space that hand was placed in the congenitally blind group. As shown in Fig. 2, this pattern of results was highly consistent across all participants within each group. Separate analyses for the four groups showed a significant interaction of Hand Posture and Spatial Correspondence only in the congenitally blind group (F_1,10 = 31.90, P < 0.001). Interestingly, whereas the sighted controls for the congenitally blind performed worse overall with their hands crossed than with their hands uncrossed (Hand Posture: F_1,10 = 11.34, P = 0.007), the congenitally blind were not significantly affected by hand crossing (P > 0.18). A similar main effect of Hand Posture was also observed for the late blind and their matched sighted group (F_1,12 = 6.75, P = 0.023). Mean error rates were below 10% in all groups and conditions. The same pattern of results as for the IE scores was obtained for RTs [see supporting information (SI) Fig. 3] and error rates (see SI Fig. 4).

Fig. 1.

Group mean IE scores with standard errors of the means in Experiments I (Simon task, a and c) and Experiment II (b and d) for the congenitally blind and their control group (a and b) and the late blind and their control group (c and d). IE scores are calculated by dividing RTs by correct response rates, thus eliminating any potential speed-accuracy tradeoff effects in the data; the lower the score, the more efficient/better is performance (i.e., analogous to the RT data) (19, 36). Experiment I (a and c): With hands uncrossed, all groups showed a similar pattern of inferior performance in the noncorresponding compared with the corresponding condition. This effect reversed in the congenitally blind only when their hands were crossed (with respect to an external coordinate system). All of these groups, except the congenitally blind, activated an external reference frame by default. Experiment II (b and d): When the congenitally blind had to match sound and hand locations, interference due to hand crossing was larger than in the remaining groups. It is possible that only the congenitally blind had to inhibit their default anatomical reference frame and activate an external reference frame, whereas the latter had automatically been activated by default in the remaining groups. Note the different scales.

Table 1.

Results of the overall ANOVA: Visual Status (blind vs. sighted) × Hand Posture (uncrossed vs. crossed) × Spatial Correspondence (corresponding vs. noncorresponding) for inverse efficiency scores as the dependent variable

Effect	df1/df2	F	P
Congenitally blind vs. matched sighted
VS	1/20	0.41	n.s.
P	1/20	0.90	n.s.
SC	1/20	61.82	<0.001
VS × P	1/20	10.09	0.005
VS × SC	1/20	28.92	<0.001
P × SC	1/20	27.07	<0.001
VS × P × SC	1/20	19.63	<0.001
Late blind vs. matched sighted
VS	1/12	1.12	n.s.
P	1/12	6.75	0.023
SC	1/12	41.47	<0.001
VS × P	1/12	0.06	n.s.
VS × SC	1/12	0.02	n.s.
P × SC	1/12	0.14	n.s.
VS × P × SC	1/12	3.67	>0.08

VS, Visual Status; P, Posture; SC, Spatial Correspondence; n.s., not significant; df1, degrees of freedom in the numerator; df2, degrees of freedom in the denominator.

Fig. 2.

Experiment I: Inverse efficiency scores for individual participants: sighted controls for the congenitally blind (a), the congenitally blind (b), sighted controls for the late blind (c), and the late blind (d). Note that only the congenitally blind participants showed a reversed Simon effect with their hands crossed. Note the different scales for a and b vs. c and d. Late blind participants 01 and 02 had been totally blind for 34 and 50 years, respectively.

Many researchers have shown that the Simon effect is susceptible to sequential effects, that is, performance in noncorresponding trials is inferior to that seen in corresponding trials only if the preceding trial was of the corresponding type. This pattern of results was also observable in the two sighted and the late blind groups, irrespective of hand posture. Interestingly, however, the sequence effect was reversed in the congenitally blind group when their hands were crossed: that is, they performed worse in the corresponding trials than in the noncorresponding trials after noncorresponding but not after corresponding trials (see SI Figs. 5 and 6). This finding corroborates the assumption that stimulus-response correspondence in this group of participants was determined by the anatomical location of the hands rather than by their externally anchored spatial position.

In sum, the results of Experiment I confirmed our hypotheses: Whereas sighted and late blind adults used a stimulus-location/response-location mapping, i.e., an external coordinate system as their default, the congenitally blind displayed a bias to invoke a stimulus-location/response-hand mapping instead, i.e., an anatomically anchored reference frame to control their action. This pattern of results suggests that developmental vision is required to set up a bias for using an external reference frame for auditory-manual control.

If the congenitally blind participants were the only ones to use a stimulus-location/response-hand (rather than a stimulus-location/response-location) mapping as their default for auditory-manual control, one would have expected their performance to be inferior to that of the sighted and late blind individuals in any task that explicitly required the mapping of the location of the target sound and the responding hand. This hypothesis was tested in Experiment II, which differed from Experiment I only in the instructions given to participants (i.e., all other aspects of the design were identical): The participants were assigned to respond with the hand that was located on the same side as the sound (irrespective of the pitch of that sound) while their hands were held in either an uncrossed or a crossed posture. Accordingly, and in contrast to Experiment I, this task required an explicit matching of the spatial location of the sound with the spatial locations of the hands while keeping stimulation and the experimental setting the same.

Experiment II.

Whereas no significant differences were found between the late blind and their age-matched controls (Fig. 1d), the performance of the congenitally blind was worse than that of their controls when they completed the task with their hands crossed (see Fig. 1b). In other words, hand crossing impaired the performance of all groups, but did so more in the congenitally blind (see Fig. 1 b and d) (two-way interaction between Visual Status and Hand Posture: Congenitally Blind vs. Sighted: F_1,20 = 9.77, P = 0.005, with one-tailed posthoc t test Congenitally Blind vs. Sighted in Crossed Hands Condition, P = 0.0395; none of these effects reached significance in the analyses comparing the late blind to their matched sighted controls). Error rates were below 10% in all groups and conditions. A corresponding pattern of results as for the IE scores was observed for both the RTs and error rates (see SI Figs. 3 and 4, respectively).

It has been suggested on the basis of neural modeling studies (5) that the use of external coordinates for action control might be advantageous even if modality-specific reference frames are sufficient for solving a particular task, as in the case of auditory-manual control. Our study provides compelling support for this hypothesis in living systems: When the congenitally blind participants had to remap the auditory inputs into an external reference system, they showed a significantly larger performance decrement than the sighted and the late blind, thus suggesting that the latter groups have an advantage in terms of their default use of external frames of reference.

Discussion

The aim of the present study was to investigate whether the use of external coordinates for multisensory action control is acquired or innate, and whether it is mediated by the increasing dominance of vision in manual control during ontogeny. Moreover, we tested whether the use of a nonanatomical coordinate system can be beneficial for nonvisual auditory-manual control, i.e., in tasks that could (theoretically) be performed without remapping the action goal into external coordinates. In an auditory version of the Simon task (Experiment I), we found a reversed Simon (i.e., spatial compatibility) effect in the congenitally blind when they performed the task with their hands crossed, suggesting that, by default, they activate an anatomical coordinate system, whereas the sighted and late blind groups mapped spatial information with respect to an external reference frame. When, as in Experiment II, the congenitally blind were forced to match the locations of the target sound and their responses, they showed a significantly larger performance decrement than the sighted and the late blind groups when adopting a crossed hands posture, i.e., when anatomical and external reference frames were placed in conflict. This double dissociation of detrimental effects due to hand crossing in the two experiments (i.e., larger detrimental effects of hand crossing in the sighted and late blind than in the congenitally blind in Experiment I (Simon task), but larger detrimental effects in the congenitally blind than in the sighted and late blind in Experiment II) clearly demonstrates that the default remapping of auditory and proprioceptive inputs into external coordinates is acquired during development as a consequence of visual input. Such a double dissociation also allows us to exclude “unspecific” accounts, such as general task difficulty or inhibitory effects for the group differences that we observed. Moreover, our results show that the default remapping of sensory inputs into a modality-invariant coordinate system is indeed of advantage for action control even toward nonvisual events, that is, events which do not necessarily require the activation of an external coordinate system. Our data are therefore consistent with the hypotheses that have emerged from neural modeling approaches (5) suggesting that sensory stimuli are coded in external or visual (eye-centered) coordinate systems (see footnote on page 1). Neural network models have shown that the use of such a common frame of reference facilitates performance in many different tasks, including motor planning to objects irrespective of their modality, and also predicting the consequences of a motor action on sensory input from different modalities. Moreover, an inherent feature of these networks is that certain neurons display partially shifting receptive fields (RFs) which have frequently been observed in the posterior parietal cortex (20, 21) and the superior colliculus (22). For example, single cell recordings have recently provided evidence of auditorily responsive neurons in the ventral intraparietal area (21). These neurons were either bimodal (visual and auditory) or even trimodal (visual, auditory, and tactile). A subpopulation of the auditory-visual neurons had overlapping RFs and coded space according to a common reference frame (21). Furthermore, Schlack et al. found neurons with partially shifting RFs, i.e., an incomplete transformation of auditory RFs into an eye-centered/external coordinate system. It has been hypothesized that partially shifting RFs are one reason for the shift in the subjective auditory median plane of the head in humans when they either turn their eyes or their head to one side (23). In line with this view, deactivation of the parietal cortex by using TMS appears to eliminate this effect (24). Importantly, congenitally blind individuals do not show such a shift of the subjective auditory median plane when turning their head (25). Thus, unlike sighted individuals, the congenitally blind do not seem to use the neural systems that, in sighted individuals, use proprioceptive inputs to provide a default remapping of auditory space into external coordinates.

Interestingly, in the present Experiment I, the congenitally blind were the only group who did not perform worse overall in the crossed hands condition. On the one hand, this finding implies that the overall worse performance of the sighted and late blind adults when their hands were crossed may not purely be of motoric origin but may instead reflect a conflict between anatomically and externally defined coordinate systems (see 8). As in tactile localization, sighted (and late blind) individuals do not only seem to activate a single frame of reference at a time for auditory manual control. Rather, they appear to activate several reference frames in parallel, consistent with recent suggestions based on animal research (26). On the other hand, the lack of an overall performance decrement in the congenitally blind group in Experiment I (Simon task) corroborates the hypothesis that the larger performance decrement observed for the congenitally blind than for the sighted due to hand crossing in Experiment II resulted from a more time-consuming remapping process of sound and hand location rather than from simple peripheral (e.g., motor) processes.

When taken together, the findings of Experiment I and II would appear to indicate that the mapping of sensory inputs into external coordinates is not innate, but occurs as a consequence of visual input during ontogeny. Visual inputs seem to act as the driving force for setting up this common external reference frame for multisensory integration and action control. This view is supported by the fact that some auditory spatial maps are shaped by visual input during development (11). Wallace et al. showed that although cats and monkeys are born with neurons that respond to stimuli in more than one sensory modality, the specific multisensory integration properties, expressed in a superadditive response when both modality components originated from the same location, were not innate but developed during the first few months of life (27). More recently, Wallace et al. (28) have reported studies in which they raised cats with bimodal stimuli with the auditory and visual component always presented from different locations. These animals started to integrate these deviant pairings (at least to some extent) suggesting, in accordance with earlier findings (11), that the specific spatial mapping of input of different senses is not innate either. Finally, Wallace et al. demonstrated that animals raised in darkness have multisensory neurons but that they do not show any sign of multisensory facilitation for spatially corresponding stimuli, either for any pairings with visual stimuli or for auditory-somatosensory bimodal stimuli (29). These findings are consistent with an event-related potential study in congenitally blind humans demonstrating that the congenitally blind do not show cross-talk between tactile and auditory systems based on spatial features at the level of early sensory processing (12). Together with the present findings, these reports suggest that the default or automatic remapping of sensory inputs into a common external frame requires visual input.

The data reported here are also consistent with earlier reports of a failure to remap tactile stimuli into an external reference frame in the congenitally blind (13). However, whereas these previous findings focused on perceptual processes, the present study extends these earlier results to the case of action control. To our knowledge, however, the behavioral consequences of the lack of developmental vision on multisensory action control have not been investigated in animals to date.

It could be argued that the default use of an external reference frame for manual control is innate and that congenitally blind individuals switch to the use of an anatomical reference frame because of their sensory impairment rather than the sighted acquiring the default use of an external reference frame during development. However, this explanation appears unlikely. Prospective cross-sectional developmental studies have shown that the interfering influence of conflicting proprioceptive information on auditory localization decreases during childhood and that auditory information increasingly affects proprioceptive information with increasing age (30). Interestingly, a similar trend was not observed in congenitally blind individuals, thus suggesting that vision does indeed change both auditory and tactile/proprioceptive localization.

The data reported here do not allow one to pinpoint exactly when during ontogeny external reference frames start to become dominant in multisensory action control. Based on the results of the late blind in the present study and in an earlier experiment (13), it is nevertheless possible to conclude that this switch takes place before 12 years of age. The finding that the late blind do not differ from the sighted in the default coordinate systems used for auditory-manual control and tactile localization (13) suggests that the brain systems that mediate multisensory integration and action control do not remain plastic throughout life. Developmental visual input may trigger selective and constructive developmental mechanisms that irreversibly result in the coding of spatial information within an external reference frame to facilitate multisensory integration (13, 14), even when an individual has been blind for several decades (see, for example, late blind participant 02 in Fig. 2, who had been blind for 50 years).

Animal studies have shown a tremendous decrease in the number of visually responsive neurons in parietal cortex after congenital visual deprivation (31). Carlson (32) followed up on these monkeys after their eyes had been opened for a year. Interestingly, despite “normal” visual input, the trend toward less visual responsiveness in parietal cortex was found to have increased still further rather than have reversed, as one might have expected. One speculation is that visual input in these animals was no longer capable of changing the response properties of auditory and tactile neurons, just as seen after monocular deprivation where the deprived eye has been shown to be incapable of regaining synaptic space in visual cortex (33), and, as a consequence, binocular neurons failed to develop. One consequence of the decreased plasticity of multisensory brain structures in later life might be the lack of visual functional recovery in humans who were blinded early on in life and whose sight was restored when they were adults. They have been found to be particularly impaired in “dorsal” functions, such as spatial attention control and object localization (34).

Nevertheless, despite the fact that auditory-manual control seems to be facilitated by visual input during ontogeny, auditory spatial maps do develop without vision, and even, to some extent, with a higher precision (35). It is important to note that the congenitally blind performed well in the task used in Experiment II, i.e., their mean error rate was below 10%. Moreover, all of the participants, including the sighted and late blind, displayed some performance decrement when they had to match the hand and sound locations with their hands crossed (as in Experiment II). Therefore, anatomical and external coordinate systems appear to be set up and accessible regardless of the availability of visual inputs. However, both the reversed Simon effect seen in the congenitally blind group in Experiment I, as well as the larger performance decrement reported in this group in Experiment II due to hand crossing (and presumably due to a conflict between anatomical and external reference frames) suggest that the choice of the default reference frame for perception and action control as external or anatomical is defined by visual experience during ontogeny.

Materials and Methods

Experiment I.

Participants.

Two groups of blind (one congenitally blind and one late blind) and two groups of sighted adults took part in both Experiments I and II.

1. The congenitally blind group consisted of 12 adults of whom one was unable to comply with the task instructions and was therefore not included in the sample. The remaining congenitally blind participants (4 female, mean age: 35.6 years, range: 23–47 years, 9 right-handed, 2 bimanual) were totally blind (n = 8) or else did not have more than rudimentary sensitivity for brightness differences without any pattern vision. Blindness was due to peripheral reasons in all cases (for details see Table 2).

2. The sighted control group for the congenitally blind was matched in age, gender, and handedness and consisted of eleven adults with normal (n = 2) or corrected-to-normal vision (4 female, mean age: 35.2 years, range: 23–51 years).

3. The late blind group comprised seven right-handed adults (all female, mean age: 55.9 years, range: 37–74 years). Blindness was due to peripheral causes and either total (n = 4) or else left no more than rudimentary sensitivity for brightness differences again without pattern vision. The onset of blindness occurred after the age of 15 years (mean age: 34.1 years, range: 15–58 years) and the minimum duration of blindness before participation in the study was ten years (mean: 21.7 years, range: 10–50 years).

4. The sighted control group for the late blind comprised 7 right-handed women with normal (n = 1) or corrected-to-normal vision (mean age: 56.6 years, range: 38–72 years). Two of the sighted participants served as controls for both the congenitally blind and late blind participants.

Table 2.

Description of the congenitally and the late blind participants

Participant	Visual perception	Age of onset (years)	Duration (years)	Age (years)	Cause of blindness	Education
CB1	None	Birth		27	Retinoblastoma	High school
CB2	None	Birth		43	Retinopathy of prematurity	High school
CB3	None	Prenatal		28	Optic nerve atrophy	High school
CB4	Diffuse light	Birth		29	Perinatal hypoxia	High school
CB5	Diffuse light	Prenatal		47	Optic nerve atrophy	Middle school
CB6	None	Birth		44	Retinopathy of prematurity	University
CB7	None	Birth		28	Retinoblastoma	High school
CB8	Diffuse light	Prenatal		43	Retina degeneration	High school
CB9	None	Birth		47	Retinoblastoma	Middle school
CB10	None	Birth		23	Retinopathy of prematurity	Middle school
CB11	None	Birth		33	Retinopathy of prematurity	High school
LB1	Diffuse light	40	34	74	Pigmentary retinopathy	High school
LB2	None	15	50	65	Glaucoma	Middle school
LB3	Diffuse light	50	11	61	Pigmentary retinopathy	University
LB4	None	21	18	39	Diabetic retinopathy	High school
LB5	Diffuse light	34	10	44	Diabetic retinopathy	High school
LB6	None	58	13	71	Pigmentary retinopathy	Middle school
LB7	None	21	16	37	Glaucoma	Middle school

CB, congenitally blind; LB, late blind.

The blind participants and the matched sighted controls were recruited from the local community and from the University of Hamburg. All of the participants reported normal hearing and no history of neurological illness. They were naive as to the aim of the study and were compensated monetarily. Some of the matched sighted participants chose to receive course credit. All of the participants gave their informed consent. The study was conducted in accordance with the ethical standards laid down in the Declaration of Helsinki (2000).

Procedure.

The participants were blindfolded throughout the experiment. They were seated at a table with their arms resting on the tabletop, facing straight ahead. The participants' head was fixed by means of an adjustable chin rest. Each trial started with a 1,000-Hz warning tone [duration: 250 ms, including a rise and fall time of 10 ms each; 73 dB(A)] presented simultaneously from two loudspeakers (Multimedia speakers A4 Tech; Simon) located 50 cm from the participants and separated by 120 cm. The target stimulus followed 800–1,000 ms (uniform distribution) after the warning stimulus. Two different bursts of pink noise [target 1: 500–5,000 Hz, 71 dB(A); target 2: 500–15,000 Hz, 74 dB(A)] served as the target stimuli (duration: 100 ms, including 10 ms rise and fall times). Target sounds were presented randomly, either from the left or right loudspeaker. The intensity of the two different targets was adjusted so that they were subjectively perceived as being equally loud (as determined in a pilot study). The participants had to decide within 2,500 ms (from the onset of the target) whether the higher or lower sounding pink noise burst had been presented (irrespective of its location); otherwise the trial was terminated. Responses consisted of lifting either the left or the right index finger out of a light key situated on the table, and separated by 50 cm. The sound/response-location assignment was counterbalanced across participants. The next trial started 1,000–1,200 ms after the response on the preceding trial. The task was performed either with an uncrossed or crossed hands posture. The stimuli were presented in a pseudorandomized order in eight blocks of 60 trials each (i.e., 15 trials of each of the 4 conditions (sound [target 1 vs. target 2] × loudspeaker [left vs. right]). Hand posture was altered after every two blocks; half of the participants started with the uncrossed posture and the other half with the crossed hand posture. Before the eight experimental blocks, each participant completed two blocks of practice trials, one with their hands uncrossed and the other with their hands crossed.

Data Analysis.

Trials with response times faster than 150 ms <1% of trials in all analyses) or slower than 1,000 ms (<1% of trials in all analyses) were eliminated. Error rates and mean RTs (for trials with correct responses only) were calculated for each participant for each of the four conditions [Hand Posture (uncrossed vs. crossed) × Spatial Correspondence (corresponding vs. noncorresponding)]. To take both response speed and accuracy into account, IE scores (19) were derived by dividing response times (in ms) by correct response rates separately for each condition. IE scores can be considered as “corrected reaction times” that eliminate any tendency toward a speed-accuracy tradeoff (36). IE scores were submitted to a three-way analysis of variance (ANOVA) with Visual Status (sighted vs. blind) as a group factor and Hand Posture (uncrossed vs. crossed) and Spatial Correspondence (corresponding vs. noncorresponding) as repeated-measures factors. RTs and error rates were analyzed in the same way. Because corresponding results as for IE were obtained, only the latter ANOVA results are reported, but figures displaying RTs and error rates are provided as SI Figs. 3 and 4, respectively. Because of the age difference between the congenitally blind and late blind participants, separate analyses were calculated for each blind group and its control group. Higher order interactions were further explored in subordinate ANOVAs and with appropriate posthoc t tests.

Experiment II.

The order of Experiment I and II was counterbalanced across participants in each group.

Participants.

The same participants as those who took part in Experiment I.

Procedure.

Experiment II differed from Experiment I only in the task that the participants had to perform. Instead of responding to the pitch of the noise bursts they had to respond with the hand that was located on the same side as the sound while their hands were held in either an uncrossed or a crossed posture. Because there were only two conditions (Hand Posture: uncrossed vs. crossed), four instead of eight blocks of 60 trials each were run.

Data Analysis.

Same as in Experiment I. The IE scores were submitted to a two-way ANOVA with Visual Status (sighted vs. blind) as a group factor and Hand Posture (uncrossed vs. crossed) as repeated-measures factor.

Abbreviations

RT

reaction time

IE

inverse efficiency.