Recalibrating vision-for-action requires years after sight restoration from congenital blindness

Being able to perform goal-directed actions requires predictive, feed-forward control, including a mapping between the visually estimated target locations and the motor commands reaching for them. When the mapping is perturbed, e.g., due to muscle fatigue or optical distortions, we are quickly able to recalibrate the sensorimotor system to update this mapping. Here we investigated whether early visual and visuomotor experience is essential for developing the ability to recalibrate. To this end, we assessed young individuals deprived from vision due to dense congenital bilateral cataracts, who were surgically treated for sight restoration only years after birth. We compared their recalibration performance to such distortion to that of age-matched sighted controls. Their recalibration performance was impaired right after surgery. This finding cannot be explained by their still lower visual acuity, since blurring vision in controls to a matching degree did not lead to comparable behavior. Nevertheless, the recalibration ability of cataract-treated participants improved with experience, matching controls’ performance after around 2 years from surgery. Thus, the lack of early visual experience affects visuomotor recalibration. However, this ability is not lost but slowly develops after sight restoration, highlighting the importance of sensorimotor experience for brain plasticity late in life.


54
Most of our visually controlled actions, such as grasping objects, proficiently walking 55 to target locations, effortlessly making use of tools or alike, are skillful and adept. Such a fast 56 and efficient behavior is difficult to achieve relying solely on feedback control, because 57 sensory feedback during movements is typically delayed and would require constant 58 monitoring. Hence, to achieve such a proficiency, we predominantly rely on feedforward 59 control. Feedforward control avoids the use of online feedback (visual and/or proprioceptive) 60 by including accurate model predictions. For example, successfully reaching for targets 61 based on predictive feedforward control requires a model including the mapping between the 62 visually estimated target location and the motor commands necessary to reach for it (Wolpert 63 et al., 2000;. Acquiring such a model needs plenty of experience and thus time. As the 64 state of our body, the range of tools we use, or the visual input constantly change, this suffering from congenital dense bilateral cataracts, surgically treated 5-19 years after birth 114 and tested days to years after surgery, with that of two control groups (Figure 1- Table   115 supplement 1). The first control group consisted of 20 typically developing sighted 116 participants who were individually matched for age with the cataract-treated sample, as we 117 found that age has an influence on the recalibration rate in the healthy population (

130
We first verified whether the cataract-treated participants under normal conditions sensorimotor system: compared to sighted controls, the group of cataract-treated participants 143 only marginally reduced their pointing error, although the learning was still significant (i.e., the 144 95% confidence interval for recalibration rate b=-0.09 CI=[-0.14, -0.03] did not include 0).

145
Instead, sighted participants tested in normal visual conditions had the fastest learning rate

173
The analysis of irecal agrees at the group level with the previous results of the rate parameter: 174 it significantly differed across the three groups and, although irecal was substantially lower in 175 the cataract-treated participants compared with the two control groups, it was still significantly 176 greater than zero ( Figure 1B, inset).

177
As the poorer performance of cataract-treated participants cannot be explained by the 178 diminished visual acuity alone-CSF-matched controls still recalibrated significantly faster-,

179
we also explored the contribution of time and experience that could contribute to the 180 development of recalibration. To this end, we correlated the performance of the cataract-181 treated participants not only with their visual acuity, but also with the time since surgery and 182 the age at test and at surgery. irecal did not correlate with age at test or with age at surgery 183 ( Figure 1D-E). This is at odds with the performance of the typically sighted population, in 184 which we found that irecal increases with age and thus experience (Figure 1-figure supplement   185 2), and it is already greater in the youngest control children (6-7-year-old, n=11, irecal 186 =0.62±0.07) than in the whole cataract-treated sample (0.30±0.06, Wilcoxon-Mann-Whitney, 187 p=0.03). This may well indicate that learning to recalibrate is not merely an effect of brain 188 maturation related to age, but requires experience to develop. This is confirmed by the fact 189 that recalibration performance is correlated with time since surgery (Pearson's correlation 190 coefficient, r=0.59, p=0.017). Additionally, it is also correlated with visual acuity (r=0.5, 191 p=0.025, Figure 1C), meaning that higher visual acuity and thus a better visual input leading 192 to better quality experience led to better recalibration performance. As learning typically 193 follows an exponential function rather than a linear trend, we also fitted the time-since-194 surgery data using an exponential:  Figure 1F).

205
The contribution of post-surgical experience to the development of recalibration at the 206 individual level is also appreciable when considering the difference in recalibration 207 performance ( "#$%& ) between each CSF-matched pair of participants: control minus 208 cataract-treated ( Figure 1F, inset

212
The role of experience for improving the recalibration performance is further highlighted in 213 Figure 1G, which focuses on the 13 individuals repeatedly tested.

287
Here we investigated whether individuals suffering from congenital dense bilateral 288 cataracts, surgically treated years after birth, can develop the ability for visuomotor 289 recalibration, which is an essential behavior enabling efficient interaction with the world. We 290 used prism goggles to distort the visuomotor mapping and compared their recalibration 291 performance to that of typically sighted individuals matched for age and visual acuity. Unlike 292 typically developing individuals, who quickly recalibrated when exposed to distortions in the improved and started to show better recalibration performance. However, it took over two 296 years for this ability to develop to levels comparable with sighted individuals.

297
It seems surprising that a flexible visuomotor mapping takes so long to mature in cataract-298 treated individuals, since being able to quickly recalibrate the sensorimotor system is 299 essential for everyday adept behavior. Indeed, a flexible sensorimotor system, capable of 300 rapid modification, grants the possibility to rely on fast, feedforward motor control when 301 interacting with the world. Healthy individuals constantly update the mapping between the 302 visually estimated location of a target and the motor command required for reaching it. They 303 do so for instance when using tools that artificially elongate arm length, when the body itself 304 is altered while carrying objects, or when using glasses that induce optical distortions.

305
However, they also constantly refine their mapping in the presence of noise when no actual

316
From a comparison between the cataract-treated individuals and the control groups 317 we can conclude that it is neither age nor the improvement in visual acuity that is the sole 318 determining factor for the development of the recalibration performance, but that such a 319 development requires experience from interaction with the world. Firstly, we can determine 320 that age is not the sole decisive factor from comparing our cataract-treated individual to the 321 group of age-matched sighted controls, which shows that cataract-treated participants are on 322 average much less efficient in reducing the error when exposed to a prismatic shift, and they 323 also present less of an aftereffect following prism removal. In addition, the average rate of age. Furthermore, in the cataract-treated individuals there was no significant correlation 327 between the recalibration ability and age. Secondly, we ruled out that the post-surgical visual 328 acuity, which is still lower than that of controls even after cataract removal (cf. Ganesh et al.,

393
However, the fact that cataract-treated participants learn to integrate visual and 394 haptic information in a similar temporal window as they learn to recalibrate the sensorimotor 395 system is not enough to conclude that these two abilities are related. Indeed, although they 396 may share computational analogies (i.e., solving the correspondence problem), the two 397 abilities differ in a crucial aspect: while multisensory integration requires combining sensory 398 cues (e.g., visual and haptic), recalibration relies on the ability to combine one or more 399 sensory cues to motor commands. In other words, while the former is a purely sensory 400 process, the latter is a sensorimotor skill, thereby the two abilities rely on distinct neural     Table   477 supplement 1).

478
We had the chance to test 8 out of the 20 cataract-treated individuals also prior to 479 their surgery. Among them, only 4 had enough residual vision to be able to see the target and 480 perform the task, and were therefore tested (mean age: 12 years, range: 8-15 years, mean 481 pre-surgical visual acuity: 2.14 cpd, range: 1.31-2.91 cpd). We retested these 4 participants,

482
as well as further 9 individuals from the set described above (i.e., a total of 13 participants)

500
We ascertained that typically sighted German controls would not differ from typically

535
Visual assessment in cataract-treated participants. The vision of the participants suffering 536 from congenital bilateral cataract was evaluated prior to treatment and after cataract removal 537 surgery (see Table supplement

551
Participants rested their head on a chin-rest at 30 cm distance from the display and had to 552 report whether the grating was oriented horizontally or vertically on each trial. Some 553 participants with extremely poor vision were allowed to perform the test at a shorter distance

554
(15-20 cm) since they would have not been able to perform the task otherwise. In a first 555 block, gratings were all presented at 100% contrast. The test started with a grating at the 556 lowest spatial frequency (0.042 cpd = 1 cycle per 512 pixels at 30 cm viewing distance). As 557 long as the participant's response was correct, a grating with the next higher spatial

627
In this pre-prism phase, participants wore plastic goggles without any distorting lens and thus 628 they had a natural view on the scene. In the next phase (prism), a prismatic lens was

654
Note that the individual profiles in sighted participants were best described by exponentials.

655
However, due to noise in the individual profiles of the cataract-treated participants, the curves 656 were fitted on the group mean, and the mean of multiple exponentials with different rate 657 parameters typically approximates a power function.

658
To

941
(measured as contrast sensitivity function (CSF) cutoff frequency in cycles per degree (cpd)) 942 before and after surgery, and time since surgery at test (years (y), months (m), days (d)). In 943 the pre-surgical visual assessment, LP indicates only light perception, followed by hand surgery, either because they had too poor visual acuity to be able to perform the CSF test, or 948 because the procedure was not available at the time they were surgically treated. We 949 assessed the post-surgical CSF cutoff frequency in the same experimental session as the 950 experimental task. Some participants were tested in the study multiple times, before and/or 951 after surgery. We do not have information regarding the exact date of surgery of two 952 participants, because they were included in our project only after surgery, and not operated 953 by our team. Both of them were surgically treated more than 2 years before taking part in the 954 present experiment.