Stereopsis refers to the ability for the brain to resolve slight differences in the projections of images onto the left and right retinas, in order to form a representation of depth. Stereoscopic vision is valuable for interacting with the environment, allowing individuals to distinguish objects, estimate distances, and manipulate them effectively. As such, deficits in this capacity, as can be seen in individuals with binocular dysfunction (e.g., those with amblyopia (Levi et al., 2015)), may lead to struggles with every-day visuomotor tasks.

In recent years, an ever-growing body of work has focused on developing methods that promote neuroplastic changes in the visual system. In this vein, one technique gaining traction is brain stimulation. Recent studies have demonstrated that offline repetitive transcranial magnetic stimulation (rTMS) over the primary visual cortex (V1) can facilitate a wide range of visual functions, including contrast sensitivity and orientation discrimination (Thompson et al., 2008; Waterston and Pack, 2010). Since V1 presents the first emergence of binocular neurons, stimulation over V1 may also facilitate the analysis of depth positions. Indeed, Tuna et al. (2020) have shown that continuous theta burst stimulation (cTBS; Huang et al., 2005) over V1, despite its posited inhibitory effect on cortical excitability, improves a wide range of visual capacities in amblyopic adults, including visual acuity, suppressive imbalance, and most notably, stereoacuity.

Whether the beneficial effects of cTBS on stereovision are restricted to stimulation over V1 is unknown. There is no particular reason to assume that V1 stimulation should produce the most optimal benefits to stereopsis. Neurophysiological studies have shown that the computation of binocular disparity begins from V1 and extends to dorsal and ventral visual regions that are engaged in a task-dependent manner (Preston et al., 2008; Tsao et al., 2003; Uka and DeAngelis, 2006). Specifically, ventral areas including V4, lateral occipital complex (LOC), and inferior temporal cortex (IT), are engaged during fine discriminations of stereo-features (Preston et al., 2008; Shiozaki et al., 2012; Uka et al., 2005). In particular, previous work has shown that LOC activation reflects the categorical distinction between “near” and “far” depth positions in a fine depth discrimination task (Preston et al., 2008). Human lesion studies also revealed that cortical damage in the LOC interrupts fine disparity discrimination (Read et al., 2010). Using a disruptive online rTMS protocol, Chang et al. (2014) similarly revealed that LOC stimulation selectively hampers fine stereoscopic performance in healthy vision, and that its relevance changes following perceptual training. Given the prominent involvement of LOC in fine depth judgments, it might yet be a more sensible locus to modulate in an attempt to enhance stereovision.

We investigated here whether cTBS over LOC can bring about improvements in stereoscopic depth perception. Specifically, we tested the specificity of cTBS-induced effects (if any) to stimulation site and to visual feature (are improvements depth-specific?). Establishing the feature/task-specificity of cTBS-induced effects at the various stimulation sites will be particularly revealing given previous work showing the breadth of improvements attainable following V1 stimulation (Tuna et al., 2020), at least in Amblyopes. We expect, on the contrary, that ventral stimulation may not produce the same generalized improvements across features given its putative specialization for form and shape, but also to stereo (Chang et al., 2014; Grill-Spector et al., 2001; Kourtzi et al., 2003).

We tested three stimulation sites (early occipital cortex, V1/V2; LOC; and control site vertex, Cz). We also tested two tasks: a fine stereo-depth discrimination task and a secondary luminance discrimination task. While in principle, we could have selected any secondary feature in order to probe the specificity of cTBS-induced effects at our stimulation sites, we opted for a luminance discrimination task as it carries a feature generally processed very early in the visual cascade (Goodyear and Menon, 1998; Albrecht and Hamilton, 1982), and can be judged devoid of fine form or features that would otherwise engage well-established lateral-occipital areas (Avidan et al., 2002; Grill-Spector et al., 1999; Malach et al., 1995). We compared performances on depth and luminance discrimination tasks before and after stimulation over early occipital cortex (V1/V2), LOC, and control site vertex (Cz). In the depth task, participants were asked to discriminate between small depth positions (Chang et al., 2014; Chang et al., 2013). In the luminance task, stimuli were designed analogously to those in the depth task, but participants were instructed to ignore the depth positions and to instead judge the luminance differences in the stimulus.

Thresholds for the depth and luminance tasks are presented for the different stimulation groups and across the pre and post-tests in Figure 3A. To facilitate visual comparisons, the data were replotted in terms of normalized thresholds (pre-post/pre) in Figure 3B. See also Supplementary Figure 1 for an overlay of individual subject performance, before and after stimulation, for both the depth and luminance tasks.

To rule out the possibility that changes in performance, if any, were caused by differences in baseline performance among groups, we first compared pre-stimulation thresholds across the stimulation groups. Two one-way ANOVAs (one for each task) showed that there was no significant difference in baseline disparity [F_(2,49) = 0.006, p = 0.994, η²_p < 0.001] or luminance discrimination thresholds [F_(2,49) = 0.76, p = 0.473, η²_p = 0.03] across the stimulation groups.

We then assessed the difference in disparity discrimination thresholds before and after stimulation across groups by means of a 3 (group) × 2 (test: pre versus post) mixed ANOVA. The analysis indicated no significant main effects of group [F_(2,49) = 0.813, p = 0.449, η²_p = 0.032] and test [F_(1,49) = 0.021, p = 0.884, η²_p < 0.001], but a significant group × test interaction [F_(2,49) = 5.131, p = 0.009, η²_p = 0.173]. Bonferroni corrected paired t-tests revealed that depth thresholds significantly improved (i.e., dropped) after LOC stimulation [t₍₁₇₎ = 2.66, p = 0.0166], but not after occipital [t₍₁₆₎ = 1.48, p = 0.158] or Cz stimulation [t₍₁₆₎ = 0.63, p = 0.538]. In particular, LOC stimulation improved depth discrimination thresholds by 20.5% (from 44.28 to 35.19 arcsec) at the group level (Figure 3).

We further compared performance changes induced by left vs. right LOC stimulation. Although the average improvement in disparity discrimination thresholds following right LOC stimulation (21.7%) was slightly larger than that following left LOC stimulation (19.3%), a 2 (hemisphere) × 2 (test) mixed ANOVA only indicated a significant main effect of test [F_(1,16) = 6.682, p = 0.020, η²_p = 0.295]. There was no significant effect of hemisphere [F_(1,16) = 0.303, p = 0.590, η²_p = 0.019] nor hemisphere × test interaction [F_(1,16) = 0.094, p = 0.763, η²_p = 006]. That is, improvements following cTBS over left and right LOC did not differ significantly.

Luminance discrimination thresholds were entered into a separate 3 (group) × 2 (test) mixed ANOVA. The results revealed no significant difference in luminance thresholds across groups [F_(2,49) = 0.868, p = 0.426, η²_p = 0.034] and tests [F_(1,49) = 1.905, p = 0.174, η²_p = 0.037]. There was also no significant group × test interaction, F_(2,49) = 0.400, p = 0.672, η²_p = 0.016 (Figure 3).

We tested participants’ performance in two tasks carrying analogous stimuli, but differing task demands, before and after cTBS over V1/V2, LOC, or Cz. We found that stimulation over LOC, but not over occipital cortex nor control site Cz, improved participants’ capacity to discriminate between depth positions (“near” or “far”). Stimulation-induced improvements in depth discrimination did not differ depending on the laterality of brain stimulation, although right LOC stimulation tended to induce greater improvements. These findings are in line with previous studies that have identified critical computations in LOC during depth judgments—namely, reading-out and grouping low-level visual features, representing global object structure (Grill-Spector et al., 2001; Murray et al., 2002), and discriminating near-far depth positions (Preston et al., 2008; Read et al., 2010; Chang et al., 2014).

In contrast to the changes observed for the depth task, we found that stimulation over all three sites did not elicit reliable changes in performances on the luminance discrimination task. Hence, the perceptual improvements observed following cTBS over LOC are specific to stereoscopic depth perception and do not generalize to the perception of other visual features (or at least, to the luminance feature tested here). These particular findings are in alignment with previous work that have demonstrated that LOC activations are invariant to low-level visual features and have a low contrast dependence (Avidan et al., 2002; Grill-Spector et al., 1999). That is, changes in luminance contrast do not affect the representation of object shape and identity (i.e., luminance and contrast-invariant).

An obvious curious question that follows is: how is cTBS driving enhancements in depth perception? Psychophysical performance hinges on the interplay between the sensory signals and background neuronal noise. While cTBS-induced cortical inhibition may decrease the overall activity of cortical neurons, the suppression of neuronal noise could offset the reduced signal-related activity (Allen et al., 2014). As noise arises from random fluctuations in neuronal activity (Kohn et al., 2016; Tolhurst et al., 1983), a moderate increment in inhibition (delivered via cTBS) could significantly reduce the likelihood of spontaneous noisy discharge of neurons that do not represent any meaningful information (Allen et al., 2014). Alternatively, signal-related activity primarily stems from post-synaptic potentials, rendering it relatively less susceptible to cTBS-induced cortical inhibition. When neuronal noise exhibits correlated variability in firing rates from pairs of neurons, perceptual processing faces increased difficulty in distinguishing signal from noise (Zohary et al., 1994). Improvements following cTBS observed in previous studies have been attributed to noise decorrelation (Waterston and Pack, 2010; Clavagnier et al., 2013; Hamidi et al., 2009). In particular, a model put forth by Waterston and Pack (2010) provides a theoretical framework that demonstrates how noise decorrelation could outweigh the reduced firing rate due to inhibitory cTBS, ultimately leading to a better signal-to-noise ratio and more precise processing of sensory signals. In line with this, we conjecture that cTBS over LOC may enhance depth performance here by improving representations of the disparity-defined object shape (i.e., promoting readout) through noise suppression.

The specificity of cTBS-induced improvements to stimulation site observed here suggests further that cTBS-induced noise-reduction is region-specific. Previous work probing the effects of perceptual learning and attention have reported that their behavioral consequences on noise decorrelations vary across regions. Although both learning and attention have often been found to reduce noise correlations, noise decorrelations in different areas do not always benefit population coding efficiency (Sanayei et al., 2018). For example, training-induced and attention-spotlit noise decorrelations do not benefit fine orientation discriminations in V1 nor in the dorsal medial superior temporal area (MSTd; Gu et al., 2011; Ni et al., 2018). However, they correlate well with improvements in orientation and contrast discriminations in V4 (Sanayei et al., 2018; Ni et al., 2018). Despite the obvious differences in task and paradigms between those studies and ours, we observed similarly that cTBS over ventral area LOC, but not occipital cortex (V1/V2), enhanced depth discrimination. As such, it is possible that the noise decorrelations in LOC exclusively augment the population coding efficiency pertaining to depth discrimination.

The lack of consequences observed with occipital stimulation may seem surprising in light of several previous studies showing that occipital stimulation enhances various visual capacities in visually normal adults (Waterston and Pack, 2010; Allen et al., 2014; Maniglia et al., 2019; Schaeffner and Welchman, 2019). Remarkably, our results stand in contrast to previous work that reported fine stereoscopic improvements in amblyopic adults following cTBS over phosphene-defined V1 (Tuna et al., 2020). The discrepancy between these findings and those of ours suggest perhaps that the effects of occipital cTBS on stereoscopic depth perception vary with the population tested. While occipital cTBS improves stereoacuity in amblyopic vision through yet unknown mechanisms, it is apparently not effective in healthy vision. Why might this be?

One clear difference between the amblyopic and healthy visual systems concerns their baseline stereoacuity. The baseline stereoacuity levels reported by Tuna et al. (2020) were extraordinarily high (with a median value of 100 arcsec and a mean value of 258.3 arcsec) compared to the average baseline recorded in our behavioral experiment (44.7 arcsec). Thus, it seems possible that occipital stimulation, while effective, manifests more evidently in those with poorer baseline stereovision. This baseline-dependent characteristic of effects elicited by occipital stimulation has also been reported in previous studies across various visual capacities (Thompson et al., 2008; Silvanto et al., 2018).

Amblyopic vision is thought to be hindered by anomalous binocular interaction in the nervous system (among other monocular deficiencies; Holmes and Clarke, 2006). Due to anomalous visual experience in early development, the amblyopic brain chronically suppresses the visual input from the amblyopic eye (Harrad et al., 1996; Jampolsky, 1955). Along the visual hierarchy, interocular suppression first appears at or just before the emergence of binocular neurons in V1 (Baker et al., 1974; Kiorpes and McKeet, 1999). Therefore, it is entirely possible that there is a difference between amblyopes and in visually normally individuals in terms of the stage at which performance is limited: in amblyopic vision, this may well be V1, before the disparity information enters higher-order visual areas for subsequent stereoscopic analysis. In normal vision, our data appear to indicate that performance is limited well past occipital cortex, towards LOC. The relevance of ventral cortex to stereo performance in the normal visual system is supported by previous lesion and TMS evidence showing that disruption (or lesion) of LOC substantially worsened the fine disparity discrimination (Read et al., 2010; Chang et al., 2014).

Moving forward, one potential way to test the exact mechanistic changes elicited by cTBS, in particular as it pertains to noise reduction, is by examining the possibility of reversing the effects of LOC stimulation through elevating stimulation intensity. The logic is that a greater increment of inhibition may lead to more pervasive suppression which in turn disrupts the signal-readouts and thereby the perceptual judgment (Allen et al., 2014). It would be intriguing if LOC stimulation ends up producing a disruptive effect on stereopsis when stimulation intensity elevates beyond a certain point.

Moreover, although our data indicated that cTBS over LOC enhanced performance on the depth, but not luminance discrimination task, it may be beneficial to further discern whether this effect represents a general enhancement of all LOC-related capacities (Murray et al., 2002; Beauchamp, 2005) or represents a true, specific enhancement of depth perception. As noise permeates every level of information processing in the nervous system, it is possible that the reduction of noise in LOC also benefits other functions that rely highly on the LOC activations.

In summary, we showed that cTBS can improve disparity discrimination when LOC is targeted, suggesting that LOC is a critical locus for disparity processing, and is perhaps due for further attention for its neuroplastic capacities under (likely non-amblyopic-relevant) rehabilitative circumstances. Under the present conditions, individuals who instead received occipital stimulation did not show reliable changes in stereoscopic performance. This stands in contrast to previous work that has reported enhanced stereoacuity in amblyopic vision immediately following occipital stimulation. We therefore conjecture that stereoscopic depth perception is limited by mechanisms at different stages of the visual hierarchy in the amblyopic and healthy visual systems. Considering the posited inhibitory effects of cTBS, we speculate that cTBS enhances signal-to-noise ratio by suppressing internal noise in the nervous system thereby contributing to better read-outs of the disparity signal.