The purpose of the study was to investigate factors that influence our ability to localize a remembered target with no spatial reference. The main findings are as follows: 1) In the absence of visual references, the precision of localization but not the accuracy of localization worsens as the inter-stimulus interval between target presentation and target localization increases. 2) Under dichoptic conditions, when the target and response cue were presented to alternate eyes (LR and RL), subjects made greater errors in localization as compared to same-eye condition, 3) Errors in localization during the alternate-eye trials correlated with the nonius alignment measured during the trial, suggesting that subjects are unable to compensate for their underlying phoria. 4) Precision of localization during dichoptic viewing decreased as a function of inter-stimulus interval in both same-eye and alternate-eye trials with some differences for longer ISI. Below we discuss each of these findings in greater detail.
Feedback of eye position is critical for online control of eye movements and for higher order tasks such as spatial localization. In the absence of visual reference, eye position information could be provided by efference copy and proprioception signals. Poletti and colleagues (2013), investigated the relative contributions of efference copy and proprioception to spatial localization by examining the relationship between localization error and the number of intervening saccades between target and response in a paradigm broadly similar to what we used in this study. In their study, they found overall small errors in localization (horizontal: 0.14 ± 0.97°; vertical: −0.03 ± 0.39°) which did not vary with the number of intervening saccades. Although we did not measure eye movements, longer ISI in our study implies larger number of intervening saccades and therefore, our result that mean horizontal and vertical localization errors did not vary with ISI is consistent with their finding.
Using a model-based analysis, Poletti and colleagues (2013) suggested that reliance on efference copy for spatial localization would produce a linear increase in loss of precision with each intervening saccade while using proprioception would result in constant errors. Their group data were best fit with a mixed-model wherein efference copy was used initially (when number of intervening saccades are small) with increasing contribution of proprioception with larger numbers of intervening saccades. In our study, the precision of localization worsened in all subjects as ISI increased, i.e., BCEA increased as a function of ISI. This result is once again in broad agreement with data of Poletti and colleagues with some differences. In our data, we observed some subjects rely on eye position information from efference copy (linear increase with BCEA with time) while other subjects optimized combination of efference copy and proprioception (exponential rise to maximum of BCEA with time). As a group, the multiplicative increase in localization error with ISI (Fig. 4) followed the prediction of the efference copy hypothesis where each eye movement adds its error to the localization response. In our paradigm, while an individual subject could avail of a proprioception signal to aid in spatial localization, the importance of efference copy appears to outweigh that of proprioception. It is certainly possible that the localization response is specific to task or stimulus conditions and differences in the setup between that of Poletti et al. (2013) and ours can explain the differences.
We also asked subjects to perform the localization task under conditions of dichoptic viewing where trials were randomized between same-eye and alternate-eye conditions. Errors in localization for same-eye dichoptic conditions mirrored the results of monocular viewing where mean localization error remained constant and BCEA increased with inter-stimulus interval. A larger mean error in localization was however observed during alternate-eye conditions. For monocular and same-eye conditions, eye position information from the same-eye is utilized to accurately determine the spatial location of the target. Presumably, efference copy is a single conjugate signal equal to the saccade amplitude and in the case of the same-eye condition, the efference copy signal is likely sufficient to keep track of eye position during the ISI and thereafter generate an accurate estimate needed to localize the target in space. On the other hand, in the alternate-eye condition, the alternate-eye (eye that did not see the stimulus) has drifted to its phoria location during the ISI and so an efference copy based estimate of location in space would be insufficient because the phoria error is not accounted for.
Consistent with this hypothesis, the greater errors in localization correlated well with phoria recorded during the alternate-eye trials. A decorrelated noise background, such as what we used in our testing, further helps to dissociate the eyes by introducing larger phorias and we found that the shift in localization is in the same direction as the nonius alignment. Previous studies have observed that during monocular viewing [18, 19], if the right eye is occluded and the subject has an exophoria, the stimulus would be displaced to the right and vice versa if the subject has an esophoria. Furthermore, high correlation (0.58 to 0.77) between individual measures of phoria and target illusion were observed. In our study, correlations between nonius alignment and error were 0.91, suggesting subjects are unable to compensate for their phoria when attempting to localize the target in space.
Efference copy plays an important role in the online updating of eye position and proprioception contributes to the long-term calibration of eye alignment. During dichoptic viewing conditions, eye position information along with a measure of phoria deviation is critical to perform the localization task accurately. Since proprioceptive information arises directly from muscle stretch receptors, phoria should be transduced accurately via this mechanism. The appearance of localization errors that are correlated to the phoria suggests that the visual system does not effectively use stretch response information of eye position from the extraocular muscles. We therefore postulate that during binocular dissociation, efference copy (rendered erroneous because phoria is not accounted for) and not proprioceptive signals are used to locate objects in an environment with no spatial reference. This agrees with studies on individuals with strabismus [21–23] who show significantly greater errors in open loop past pointing in the same direction of the deviated eye.
We employed several strategies to ensure that experiments were conducted without any visual spatial reference. Therefore, to identify the effect of non-retinal eye position signals, we ensured that the subject had minimal visual spatial reference by presenting a full field dynamic white noise background comprising of decorrelated random dots. Moreover, the extent and curvature of the screen covered a visual angle of 180° x 90°, and targets were presented within the central 20°of screen. Since spatial reference from peripheral retina has a negligible influence on localization, we expect any peripheral retinal error signal, if available, would be mostly ineffective in providing visual reference for localization. The relatively large precision errors in localization (BCEA), especially for longer ISI, suggest that this is indeed true. We also opted not to measure eye movements since using any kind of eye tracker that was in our possession would have provided spatial cues; instead we relied on time (variable ISI) to investigate errors in localization. Further, subjects performed the task unaided or with contact lens to minimize spatial cues from the spectacle frame. This had negligible effect on perception of the cues as the dimensions, contrast and close distance from the screen ensured high visibility of the targets. The dichoptic viewing paradigm allowed us to investigate how the transfer of eye position information from one eye to the other influences’ localization. For this we performed hue matching to ensure there was no leakage, i.e., red target visible through green lens and vice versa. The luminance of the red (2cd/m2) targets was slightly lower than that of the green (3 cd/m2), but the subjects were given enough time to adapt to the filters. Moreover, the results between the monocular and dichoptic same-eye conditions were similar.