Perception of Straight Ahead With Vision, Hearing, and Proprioception by Individuals With Low Vision

Abstract

Purpose

The perception of straight ahead (PSA) is an important reference for spatial orientation. Although individuals with low vision often report difficulties in spatial tasks, it is not clear whether they face challenges in defining PSA. We asked whether low vision leads to increased biases and variabilities in visual PSA, and whether these changes, if any, would generalize to auditory PSA, due to the dominant role of vision in spatial tasks.

Methods

Thirty-five low-vision participants and 17 age-matched control participants completed visual and auditory PSA tasks in which they adjusted the location of a brief LED light or a finger snap sound from a speaker to appear to be straight ahead. To control for proprioception, participants also completed a PSA task in which they pointed to their straight ahead while blindfolded.

Results

The magnitudes of biases and variabilities in visual PSA were significantly associated with binocular acuity and contrast sensitivity, but the laterality of the visual biases was not associated with interocular asymmetries in acuity, contrast sensitivity, or lateral asymmetries in the binocular visual field. Low-vision individuals showed larger magnitudes of biases and variabilities in the auditory PSA independent of age or hearing status, possibly indicating a cross-modal impact of vision loss on sound localization. Despite the changes in visual and auditory PSA, the low-vision participants showed intact proprioception PSA.

Conclusions

The changes in visual and auditory PSA may contribute to spatial orientation challenges encountered by low-vision individuals, and the robust proprioceptive PSA may play an important compensatory role.

During everyday navigation, we constantly monitor our heading direction to remain straight ahead on a predetermined path, approach a target, or avoid potential obstacles.^1,2 Whereas past research has reported increased spatial localization and orientation difficulties in individuals with low vision,^3–6 it is not yet clear whether low vision can lead to changes in an individual's perception of straight ahead (PSA), such as consistently misperceiving the direction of targets relative to straight ahead.

Our hypothesis that visual PSA may change with low vision comes from the characteristics of functional deficits associated with low vision. Low vision is often associated with reduced acuity and contrast sensitivity leading to uncertainty in the perception of the direction of visual targets, which may increase the variability in determining whether a visual target is straight ahead.^5,7 Further, low vision is also often associated with visual field loss. Whether from pathologies such as glaucoma and retinitis pigmentosa that primarily lead to peripheral field loss or pathologies such as macular degeneration that primarily lead to central field loss, the severity of field loss in the two eyes is often unequal.⁸ Such asymmetric loss may lead to asymmetric boundaries of the binocular field during natural viewing, with more field restrictions in the right or left hemifield.⁹ As humans have the tendency to adopt a visual centering strategy to “stay in the middle” during locomotion,^10,11 asymmetric boundaries of the visual field might bias the PSA toward the side with the wider field. Central-field loss may also bring an additional source of PSA bias due to changes in gaze direction after the adoption of a preferred retinal locus (PRL)¹² left or right of a central scotoma. Artificially induced gaze change in people with normal vision using prisms has been found to induce visual PSA biases,^13,14 and natural gaze changes due to central field loss may also show a similar result.

Vision has been reported to play a dominant role in establishing PSA in people with normal vision.^13,14 If vision impairment indeed changes visual PSA as we hypothesized, then it is possible that such changes will also affect these individuals’ auditory PSA. However, people with vision impairment increasingly rely on their hearing for spatial tasks, which argues against the possible impact of vision loss on auditory PSA. This question is further complicated by the high co-occurrence of hearing impairment (dual sensory impairment [DSI]) in low-vision individuals,^15,16 as the reduced hearing sensitivity and presence of asymmetric hearing loss per se can lead to increased uncertainties and biases in auditory PSA.^17,18 When vision and hearing both bring sources of variability and biases, their joint impacts are less predictable. We previously found that, although individuals with vision impairment alone tend to rely more on their hearing, there are large individual differences in reporting vision or hearing as their primary sense in individuals with DSI.⁶

To explore the changes in visual PSA associated with vision impairment and their possible cascading impacts on auditory PSA, we asked low-vision participants to complete visual and auditory PSA tasks by adjusting brief visual and auditory targets to their perceived straight ahead. Low-vision participants were included regardless of their hearing status to be representative of the comorbidity status of this population. Proprioceptive PSA, a body-centered cue for spatial orienting, was also measured by having the participants point to their perceived straight ahead while blindfolded. A normative baseline was established by age-matched control participants with normal vision and hearing.

Methods

Participants

Fifty-four participants were recruited for this study. The recruitment sources included the Envision Vision Rehabilitation Center and the Envision Low-Vision support groups in Wichita, Kansas. The inclusion criteria were age 18 years or older, not totally blind (i.e., no form perception) or not totally deaf (i.e., no functional hearing), normal cognitive function, and no severe balance issue or residual effects of brain trauma that could affect the straight ahead task.

Cognitive status was verified by the Mini-Mental State Examination (score > 24).¹⁹ Balance was screened by a question: “Do you suffer from severe dizziness or imbalance?” History of traumatic brain injury was screened using the 3 Question Defense and Veterans Brain Injury Center (DVBIC) Traumatic Brain Injury Screening Tool.²⁰ One participant who reported severe dizziness/imbalance and one participant who reported residual effects of their traumatic brain injury were excluded from the study. The remaining 52 participants included 35 participants with LV (age range, 22–89 years; median, 66 years) and 17 participants with self-reported normal vision (age range, 21–74 years; median, 60 years). The experimental protocol was approved by the Institutional Review Board at Wichita State University, and all participants provided written informed consent before the experiment.

Vision and Hearing Screening

Participants wore their habitual corrective lenses in the vision screening. Monocular and binocular visual acuity were tested using the Lighthouse Distance Visual Acuity chart.²¹ Monocular and binocular contrast sensitivity were tested using the Pelli–Robson contrast sensitivity chart.²² Participants in the control group all had self-reported normal vision, defined as an absence of vision problems that cannot be corrected with glasses, and they had binocular acuity ranging from −0.2 to 0.3 logMAR and binocular contrast sensitivity (CS) ranging from 1.70 to 1.95 logCS. One of the participants with self-reported normal vision had an acuity of 0.3 logMAR and was removed from the analysis due to concerns about uncorrected refractive errors or unidentified pathologies. Participants in the low-vision (LV) group had binocular acuity ranging from −0.06 to 1.66 logMAR and binocular contrast sensitivity ranging from 0 logCS to 1.95 logCS. Note that some participants in the LV group had good acuity and contrast sensitivity due to diagnoses that primarily affect peripheral vision or diagnoses that affect the macular with fovea sparing.

Because of our primary interest in binocular visual field asymmetry, all participants completed a custom binocular field test to identify the left and right boundaries of their visual field. Specifically, participants were seated while facing a black screen spanning 100° horizontally and 60° vertically at a viewing distance of 1.5 meters (Fig. 1A, left panel). A laser pointer that projects red laser lines was mounted on a tripod directly behind the seated participant. The tripod enabled translation of the vertical red laser line from left to right and right to left across the field of view. During the testing, a researcher (DB) controlled the laser pointer to project a vertical line that spanned the entire height of the screen and manually scanned the line continuously from the left edge of the screen to the right edge or vice versa, endeavoring to maintain a constant speed. An estimate from eight sample videos showed that the approximate scanning speeds ranged from 3.58 to 4.62 deg/s, averaging 4.25 deg/s. Participants fixated on a central fixation dot and verbally responded to the appearance and disappearance of the red laser line as it scanned across their field of vision. The horizontal dimension of the binocular field (in degrees) was defined by the distance between the left and right boundaries, and the lateral binocular asymmetry (in degrees) was calculated as the difference between the portions of the field to the left and right of fixation (Fig. 1A, right panel). For example, a left boundary of 20° and a right boundary of 28° indicated a maximum horizontal field of 48° and an asymmetry of 8° to the right.

Figure 1.

Illustrations of the experimental paradigm. (A, left) The horizontal visual field boundary task. The participant sat and faced straight ahead while a red vertical laser line scanned through the visual field. (A, right) An example visual field with −20° boundary to the left and 28° boundary to the right, leading to an 8° asymmetry to the right. (B) The visual and auditory straight-ahead tasks measured by speaker and LED arrays. Only the horizontal speakers and LED arrays were used in this study.

Among the 35 participants with low vision, 19 had macular pathologies primarily due to macular degeneration. To examine the relation between PRL locations and visual PSA biases, these patients’ central scotoma status and PRL locations were assessed by an ophthalmologist (DCF) using a monocular scanning laser ophthalmoscope in combination with the binocular California Central Visual Field Test (CCVFT).¹² Specifically, the ophthalmologist observed the eccentric viewing pattern a patient adopted to fixate on a target of interest. If a patient with central scotoma rotated their eyes to the right to place their PRL on a target of interest, they were using a PRL to the left of the scotoma (in visual field coordinates). Because of the low incidence of lateral PRLs in our sample (see results), we have provided qualitative descriptions of the findings instead of including PRL location as a factor in our statistical analyses.

Pure-tone audiometry at 0.5, 1, 2, 4, 6, and 8 kHz was measured without hearing aids using an air conduction audiometer (240B air/bone audiometer; Amplivox, Birmingham, UK). Pure-tone average (PTA) threshold was calculated across 0.5, 1, 2, and 4K Hz in decibel hearing level (dB HL) according to the World Health Organization standards.²³ Participants were categorized as hearing impaired if one or both ears had PTA worse than 20 dB HL.²³ Hearing threshold was calculated as the average of the two ears (in dB HL),²⁴ and hearing asymmetry was calculated as the interaural differences in the hearing thresholds (in dB HL). All control participants had normal hearing confirmed by the audiometry results. Among the 35 participants with low vision, 20 were categorized as normal hearing with the worse ear hearing thresholds ranging from −5.0 to 18.8 dB HL, and 15 were categorized with hearing impairment with the worse ear hearing thresholds ranging from 22.5 to 67.5 dB HL. Five participants were hearing aid users and were asked whether they preferred to use their hearing aids in our spatial tasks. All five participants chose to use their hearing aids throughout the experiments.

We also measured the participants’ handedness using the Edinburgh Handedness Survey,²⁵ as it has been found to be associated with sound localization biases and veering biases.²⁶ Only four of the 51 participants were left handed; therefore, handedness was not included as a covariate in our analysis. Table 1 provides a summary of vision and hearing status for the normal-vision and LV groups. Supplementary Table S1 details the vision and hearing characteristics of the LV participants.

Table 1.

Summary of Vision and Hearing Status for the Normal Vision and Low-Vision Groups

	Mean ± SD
	Control (n = 16)	Low Vision (Normal Hearing) (n = 20)	Low Vision (Hearing Impairment) (n = 15)
Age (y)	47.9 ± 17.8	54.3 ± 17.8	71.7 ± 15.8
Binocular visual acuity (logMAR)	−0.07 ± 0.11	0.71 ± 0.59	0.63 ± 0.54
Better eye visual acuity (logMAR)	−0.07 ± 0.10	0.73 ± 0.61	0.64 ± 0.52
Worse eye visual acuity (logMAR)	0.02 ± 0.12	1.03 ± 0.50	0.85 ± 0.51
Binocular contrast sensitivity (logCS)	1.93 ± 0.06	1.08 ± 0.68	1.17 ± 0.56
Better eye contrast sensitivity (logCS)	1.88 ± 0.13	1.06 ± 0.67	1.08 ± 0.51
Worse eye contrast sensitivity (logCS)	1.83 ± 0.14	0.66 ± 0.62	0.53 ± 0.55
Binaural hearing threshold (dB HL)*	4.88 ± 5.86	7.91 ± 7.17	30.04 ± 10.54
Better ear hearing threshold (dB HL)	2.58 ± 4.73	5.94 ± 7.16	23.92 ± 10.57
Worse ear hearing threshold (dB HL)	7.19 ± 7.42	9.88 ± 7.55	36.17 ± 12.97

^*Binaural hearing threshold was computed as the average between the two ears.²⁴

Straight-Ahead Tasks

Visual and Auditory PSA

The PSA tasks were performed in a dark room with the light source turned off. A custom audiovisual board (Fig. 1B) was designed to measure the visual and auditory PSA, which consisted of horizontal and vertical arrays of light-emitting diode (LED) lights (12-mm diffused flat digital RGB LED pixels; Adafruit Industries, New York, NY, USA) and speakers (SC 5.9 50 mm × 90 mm speaker; Visaton, Haan, Germany). Only horizontal arrays were used in this study. The board was a square with a width of 81 cm (32 inch) positioned 1 meter away from the seated participants, spanning 45 degrees visual angle. There were 25 LED lights horizontally with 1.8° separation between adjacent LED lights, and nine speakers horizontally with 5.6° separation between adjacent speakers. The connection and programming of the LED lights and speakers were similar to those in the study by Jones et al.²⁷ Specifically, an Arduino Uno microcontroller (SmartProjects, Strambino, Italy) connected the control computer and the LED pixels. The speakers were controlled via a Saffire PRO 40 sound card (Focusrite, High Wycombe, UK), and the audio signals were amplified using Hi-Fi 2.1 stereo amplifiers (Lvpin Technology Co., Ltd., Zhejiang, China). MATLAB (MathWorks, Natick, MA, USA) was used to control the presentation of the visual and audio stimuli and register the responses.

Participants were seated in front of the audiovisual board in a dark room. They were positioned so that their physical straight ahead, defined as the point corresponding to the middle point between the two eyes, was perpendicular to the center of the board. There was no visual fixation target. In a vision trial, a LED pixel at a random position on the horizontal array was lit with red light for 500 ms. In an audition trial, a finger snap sound lasting for 500 ms with 100-ms onset and offset ramps was presented from one of the speakers on the horizontal array. The sampling rate was 44.1 kHz, and the speakers were equalized for intensity (60-dB sound pressure level [SPL]) using a sound level meter.

Participants responded to the visual or auditory stimulus using the left and right keys on a keyboard. The keys were marked with tactile stickers to assist with locating the keys in the dark. Participants were instructed to adjust the light/sound back and forth until it was centered for them on the horizontal plane. Each modality was repeated in two blocks of four trials in randomized order. The start position of the visual or auditory target was randomized in each trial. Participants could take as much time as they needed to complete each trial. No feedback was provided after each trial.

Proprioceptive Straight Ahead

The test for proprioceptive straight ahead was similar to those used in previous literature.^28,29 Participants were blindfolded and seated in front of a clear board that allowed the experimenter to mark the landmarks of interest. Participants were instructed to first bring the index finger of their dominant hand to the center of their chest and then to point straight ahead directly in front of the center of their chest and touch the board. The researcher marked the response on the board, and the participant was instructed to bring their hand back to the center of their chest. A total of eight trials were repeated. Prior to actual testing, the experimenter provided a demonstration and then asked participants to practice to ensure that they understood the task. On the same board, the actual straight ahead, again defined as the point corresponding to the middle point between the two eyes, was also marked as a reference.

Data Analysis

Data were analyzed using R (R Foundation for Statistical Computing, Vienna, Austria).³⁰ The responses in the visual, auditory, and proprioceptive PSA tasks were converted into angles in unit of degrees. The responses were then quantified by (1) bias, which is the average of the signed errors, representing both the laterality of left or right offsets in PSA from the physical straight ahead and the magnitude of such offsets (e.g., a bias of −6° indicates an overall leftward offset with a magnitude of 6°); and (2) variability, which is the standard deviation of the signed errors across the eight responses for each task, representing the response consistency.

We first compared the magnitude of the biases and variabilities in the visual, auditory, and proprioceptive PSA between the two groups (LV and control). The Shapiro–Wilk test (shapiro.test in R stats package) was used to examine the normality of the distribution of each dependent variable. When the assumption of normality was met, an analysis of variance (ANOVA) test (aov in R stats package) was conducted to examine the group difference, and F statistics were reported; otherwise, a Kruskal–Wallis test (kruskal.test in R stats package) was conducted, and H statistics were reported.

Correlation analyses were conducted to explore the associations between the visual and auditory PSA and six vision screening parameters (binocular acuity, interocular acuity difference, binocular contrast sensitivity, interocular contrast sensitivity difference, horizontal visual field, and horizontal visual field asymmetry); two hearing screening parameters (hearing threshold and hearing asymmetry), and age. Significant factors were then submitted to multiple linear regression models to further elucidate the contributions of each factor to the PSA biases and variabilities. The distributions of the residuals of the multiple linear regression models were examined using the Shapiro–Wilk test. When the normality assumption was validated, robust multiple linear regression³¹ was conducted instead (rlm function in R MASS package). The Anova function in the R car package was used to obtain the significance of each variable, and summary stats was used to obtain the coefficient for each variable.

Results

Visual Straight Ahead

We first report the results of our binocular visual field boundary test. The control group had horizontal fields that ranged from 83.38° to 100° (maximum measurable range by the custom binocular field test) and lateral binocular field asymmetry that ranged from −0.77° to 2.69°. The LV group had significantly smaller horizontal fields that ranged from 25° to 100°, H(1) = 19.19, P < 0.001, and larger magnitude of lateral binocular field asymmetry, H(1) = 18.12, P < 0.001, which ranged from −30.58° to 10.38°. The lateral binocular field asymmetry was significantly correlated with both the interocular visual acuity difference (left eye – right eye; r = 0.31, P = 0.046) and interocular contrast sensitivity difference (left eye – right eye; r = −0.32, P = 0.022). The direction of the correlation indicates that better right eye was associated with lateral binocular field asymmetry to the right side, consistent with our hypothesis that interocular difference may be contributing to lateral binocular field asymmetry.

In the visual PSA task, the control group had biases that ranged from −2.16° to 6.17°. As shown in Figure 2A, the LV group had a wider range of visual biases that ranged from −6.84° to 12.83°, doubling the range of the control group. Although the magnitude of the visual biases did not show a significant difference between the control and LV groups, H(1) = 1.61, P = 0.20, correlation analyses showed that the magnitude of the visual biases increased significantly with worse binocular acuity (r = 0.43, P = 0.002) and contrast sensitivity (r = −0.36, P = 0.009). However, the laterality of the visual biases was not significantly associated with the lateral binocular visual field asymmetry (Fig. 2B), the interocular acuity difference, or interocular contrast sensitivity difference (full correlation results are provided in Table 2).

Figure 2.

Bias and variability in the visual perceived straight ahead task. (A) Boxplots show the distributions of the visual PSA biases (in degrees) in the control and LV groups. (B) Scatterplot illustrates a lack of correlation between visual PSA biases and the lateral binocular field asymmetry. Participants with left and right eccentric viewing positions are annotated with dashed red and solid green outlines, respectively. (C) Boxplots show the distributions of the visual PSA variabilities (in degrees) in the control and LV groups. (D) Correlation between the visual PSA variability and binocular visual acuity. In the boxplots, individual data are also plotted as overlays on the boxplots. In the scatterplots, cyan circles and orange triangles represent the control and LV participants, respectively. ***P < 0.001. EVP, eccentric viewing position; PRL, preferred retinal locus.

Table 2.

Correlation Results on Visual and Auditory PSA Biases and Variabilities

	Visual Bias (Signed) (°)	Visual Variability (°)	Auditory Bias (Signed) (°)	Auditory Variability (°)
Age (y)	r = 0.05, P = 0.75	r = 0.04, P = 0.78	r = −0.10, P = 0.49	r = 0.44, P = 0.002
Binocular VA (logMAR)	r = 0.10, P = 0.50	r = 0.56, P < 0.001	r = −0.04, P = 0.76	r = 0.33, P = 0.023
Interocular VA difference (logMAR)*	r = −0.10, P = 0.52	r = 0.03, P = 0.85	r = −0.07, P = 0.69	r = 0.17, P = 0.28
Binocular CS (log CS)	r = −0.03, P = 0.82	r = −0.49, P < 0.001	r = 0.17, P = 0.23	r = −0.31, P = 0.029
Interocular CS difference (log CS)*	r = 0.06, P = 0.69	r = −0.14, P = 0.34	r = −0.11, P = 0.45	r = −0.01, P = 0.96
Binocular visual field (°)	r = 0.16, P = 0.28	r = −0.25, P = 0.08	r = 0.14, P = 0.34	r = −0.19, P = 0.20
Binocular visual field asymmetry (°)*	r = 0.04, P = 0.78	r = 0.08, P = 0.56	r = 0.07, P = 0.63	r = 0.02, P = 0.87
Binaural threshold (dB HL)	r = 0.01, P = 0.99	r = 0.19, P = 0.19	r = 0.15, P = 0.29	r = 0.51, P < 0.001
Hearing asymmetry (dB HL)*	r = 0.16, P = 0.26	r = −0.10, P = 0.93	r = −0.09, P = 0.54	r = 0.17, P = 0.23
Vision bias (°)	—	—	r = −0.34, P = 0.017	—
Vision uncertainty (°)	—	—	—	r = 0.62, P < 0.001

^*For vision and hearing asymmetry data, signed values were used in the analyses for biases and unsigned values were used in the analyses for variabilities.

Among the 19 participants with macular pathologies, five had foveal viewing with no macular scotomas, five had foveal viewing with a surrounding ring scotoma, four had eccentric viewing position to the right field (PRL to the left field), three had eccentric viewing position to the superior field (PRL to the lower field), one had eccentric viewing position to the left field (PRL to the right field), and one had eccentric viewing position to the superior-left field (PRL to the lower-right field). Although the small sample of participants with lateral eccentric viewing positions prevents statistical analysis, in Figure 2B we annotated individuals with right (triangles with solid green lines) or left (triangles with dashed red lines) eccentric viewing positions. It can be observed that the two participants with left eccentric viewing positions had larger leftward visual PSA biases than the four participants with right eccentric viewing positions.

A large impact of vision impairment was observed in the variability of visual PSA (Fig. 2C). The control group had variability that ranged from 0° to 3.62° (mean ± SD = 1.60° ± 0.90°), and the LV group had significantly larger variability that ranged from 0° to 13.53° (mean ± SD = 4.52° ± 3.75°), H(1) = 12.8, P < 0.001. Consistent with our hypothesis, the visual PSA variability was significantly correlated with both binocular acuity (r = 0.56, P < 0.001) (Fig. 2D) and contrast sensitivity (r = −0.49, P < 0.001) (Table 2), which explained 31.4% and 24.0% of the variances, respectively, in the visual PSA variability in univariate regression models. Because acuity and contrast sensitivity were correlated with each other (r = 0.90, P < 0.001), we conducted a robust multiple linear regression model on the visual PSA variabilities with both binocular acuity and contrast sensitivity as factors and found that only binocular acuity was a significant factor in this model, coefficient = 3.33, F(1, 46) = 9.04, P = 0.004.

Auditory PSA

Figure 3A illustrates the distribution of the auditory PSA biases in the control and the LV groups. The LV group was plotted as two subgroups based on their categorization of normal hearing (LV_NH) or hearing impairment (LV_HI). The control group had biases that ranged from −4.71° to 6.37°. Both LV subgroups showed wider ranges, −18.88° to 12.57° for the LV_NH group and −16.65° to 20.35° for the LV_HI group. The Kruskal–Wallis analysis on the magnitude of the auditory biases confirmed a significant difference between the control and LV groups, H(1) = 6.44, P = 0.011. However, the laterality of the auditory biases was not significantly correlated with the binaural hearing asymmetry (r = −0.09, P = 0.54) (Fig. 3B; Table 2).

Figure 3.

Bias and variability in the auditory perceived straight ahead task. (A) Boxplots show the distributions of the auditory PSA biases (in degrees) in the control, LV_NH, and LV_HI groups. (B) Scatterplot illustrates a lack of correlation between auditory PSA biases and the interaural asymmetry. (C) Boxplots show the distributions of the auditory PSA variabilities (in degrees) in the control, LV_NH, and LV_HI groups. (D) Correlation between the auditory PSA variability and hearing thresholds. In the boxplots, individual data are also plotted as overlays on the boxplots. In the scatterplots, cyan circles, orange triangles, and orange squares represent the control, LV_NH, and LV_HI participants, respectively. ***P < 0.001.

The control group showed auditory PSA variability that ranged from 1.99° to 10.39° (mean ± SD = 4.36° ± 2.74°), replicating the overall poorer auditory localization than visual localization performance in individuals with typical vision and hearing (Figs. 2C, 3C). The LV group had significantly larger auditory variabilities than the control group, H(1) = 14.45, P < 0.001, with variabilities ranging from 2.60° to 19.35° in the LV_NH subgroup (mean ± SD = 9.37° ± 5.70°) and from 3.98° to 18.53° in the LV_HI subgroup (mean ± SD = 10.68° ± 4.83°). The auditory PSA variability was significantly correlated with age (r = 0.44, P = 0.002) and binaural hearing (r = 0.51, P < 0.001); as age increased and hearing threshold worsened, the PSA variability increased (Fig. 3D; Table 2).

Cross-Modal Correlations Between Visual and Auditory PSA

The increased biases and variabilities in auditory PSA in our LV participants with normal hearing are suggestive of potential cross-modal impacts of vision status on auditory PSA. Therefore, we explored the correlations between visual PSA and auditory PSA across all participants. Significant cross-modal correlations were observed for both PSA bias (r = −0.34, P = 0.017) and variability (r = 0.62, P < 0.001) (Fig. 4; Table 2). In Figure 4A, the PSA biases show negative correlation, indicating that left visual biases were associated with right auditory biases and vice versa. In Figure 4B, the PSA variabilities show positive correlation, indicating that larger visual variabilities were associated with larger auditory variabilities. It can also be observed that vision remained as the more reliable (smaller variability) sense than hearing in defining PSA in our sample of participants (Fig. 4B).

Figure 4.

(A, B) Cross-modal correlations in PSA bias (A) and variability (B) between vision and audition. Cyan circles, orange triangles, and orange squares represent the control, LV_NH, and LV_HI participants, respectively.

To further examine the influence of vision factors on auditory PSA performance, we ran full correlation analyses between the vision screening parameters and the auditory biases and variabilities (Table 2). Auditory PSA bias was not significantly correlated with any of the vision screening parameters, leaving the measured vision PSA bias as the sole significant factor. Auditory PSA variability was significantly correlated with binocular visual acuity (r = 0.33, P = 0.023) and contrast sensitivity (r = −0.31, P = 0.029), in addition to the binaural hearing threshold and visual PSA variabilities reported above. Therefore, in a full multiple regression analysis, we entered age, hearing threshold, visual PSA variabilities, binocular acuity, and contrast sensitivity as factors for the auditory PSA variability. ANOVA analysis showed that only age, F(1, 43) = 4.68, P = 0.036, and visual PSA variabilities, F(1, 43) = 19.98, P < 0.001, were significant, which together explained 54.8% of the variance in the auditory PSA variability. Full regression results are provided in Table 3.

Table 3.

Multiple Regression Model on Auditory PSA Variability

	Coefficient [95% Confidence Interval]	Standard Error	t	P
Intercept	−1.79 [−9.45, 5.89]	3.80	−0.47	0.64
Age	0.08 [0.01, 0.16]	0.04	2.16	0.036
Binocular VA	0.92 [−3.40, 5.25]	2.14	0.43	0.67
Binocular CS	0.51 [−3.29, 4.31]	1.88	0.27	0.79
Binaural threshold	0.08 [−0.04, 0.21]	0.06	1.39	0.17
Visual PSA variability	0.81 [0.45, 1.18]	0.18	4.47	<0.001

Proprioceptive PSA

Despite these changes in visual and auditory PSA biases and variabilities, in the proprioceptive condition where the participants closed their eyes and pointed to their perceived straight ahead, all participants had small bias and variability that were close to 0° (Fig. 5). There was no group difference in the magnitude of the proprioceptive biases, H(1) = 0.12, P =0.73, or variabilities, H(1) = 0.50, P =0.48, and no cross-modal correlations between the proprioceptive PSA with the visual or auditory PSA.

Figure 5.

Bias and variability in the proprioceptive straight ahead task. (A) Boxplots show the distributions of the proprioceptive PSA biases (in degrees) in the control and LV groups. (B) Boxplots show the distributions of the proprioceptive PSA variabilities (in degrees) in the control and LV groups.

Discussion

We investigated the impact of vision impairment on low-vision participants’ PSA using their residual vision. We found that worse binocular acuity and contrast sensitivity were associated with larger magnitude of the visual PSA biases and increased visual PSA variabilities. However, we did not find any association between the laterality of the visual PSA biases with the vision asymmetries in acuity, contrast sensitivity, or visual field. We specifically hypothesized that lateral binocular visual field asymmetry might be associated with visual PSA biases due to human's intrinsic tendency to center the visual information from left and right fields.^10,11 Using a custom binocular visual field test, we indeed observed larger horizontal field asymmetry associated with interocular differences in participants with vision impairment, but it was not a significant predictor for the visual PSA biases. Although the lack of predictors for the laterality of visual PSA bias could be due to our sampling limitations, it is also possible that these results reflect individual differences in adapting to the asymmetries.

A second goal of our research was to investigate the cross-modal impact of vision PSA change on auditory PSA. In our convenience sample recruited from a low-vision clinic, 50% of the participants were categorized as hearing impaired, consistent with the previously reported prevalence of hearing loss in low-vision patients seeking rehab services.^15,16 We found significantly larger biases and increased variabilities in auditory PSA, compared with normal controls, even in low-vision participants with normal hearing. Correlation analyses further confirmed the association between visual PSA and auditory PSA for both bias and variability, which were consistent with the dominant role of vision in spatial localization and that changes in visual spatial perception could lead to recalibrations in auditory spatial perception.^7,32,33

We were surprised to observe that the visual and auditory PSA biases tended to point toward opposite directions from the physical straight ahead, which was different from the classic “ventriloquist effect,” where the auditory localization biases tend to point in the same direction as the visual target location.⁷ This difference might be due to the functional importance of maintaining straight ahead—opposite directions of visual and auditory PSA biases could be functionally preferrable, as they would average out when locating bimodal targets straight ahead with both visual and auditory input.³³

The visual PSA variability overruled other vision and hearing screening parameters in a multiple regression analysis on the auditory PSA variability. This finding indicates that it is an individual's ability to use their residual vision in spatial tasks, rather than visual acuity, contrast sensitivity, or visual field loss, per se, that best associates with auditory spatial performance. In a recent laboratory study, we found that better visual localization performance was associated with better sound localization performance.⁵ In another study, we found that individuals with self-reported higher visual spatial abilities also reported stronger vision dominance over hearing.⁶ Research findings on auditory distance perception in low-vision participants have shown that more severe forms of vision loss were associated with poorer sound distance perception.³⁴ Our current findings add to this literature by showing the correlations between visual and auditory PSA biases and variabilities in low-vision individuals.

To exclude the possible confounding factors of proprioceptive changes on the visual and auditory PSA, we directly assessed the proprioceptive PSA by having the participants point to their straight ahead while blindfolded. Despite the changes in visual and auditory PSA tasks, our sample of low-vision participants had intact proprioceptive PSA. Proprioceptive cues have been found to be beneficial for spatial updating in low-vision and blind individuals who have reduced access to visual cues.³⁵ In our sample of low-vision participants who have significant biases and variabilities in visual and auditory PSA, the intact proprioception PSA may play an important compensatory role in defining straight ahead. For example, while navigating a new space, proprioceptive cues (e.g., direction of turns) may be more robust for spatial updating than using visual and auditory landmarks for an individual with visual and auditory PSA deficits.

It is important to note the difference in the target population between our study and previous studies on PSA in individuals with hemianopia.^28,29 Hemianopia has cortical origins; it is sometimes accompanied by other higher level functional deficits such as spatial neglect, making it difficult to determine whether the changes in perceived straight ahead are visual or attentional. Patients with homonymous hemianopia without neglect symptoms were consistently found to show shifts in visual PSA toward the blind side.^36,37 Our study focused on low-vision participants with ocular pathologies, and the changes in visual PSA we observed in the current study were likely due to vision changes per se.

Although there have been few direct investigations on the visual PSA changes in people with low vision, visual biases have been reported in other measures. For example, patients with central field loss often show impairments in their hand-eye coordination in tracing and grabbing tasks,^38,39 which may be related to the discrepancy between visual and proprioceptive PSA observed in our study. Patients with retinitis pigmentosa were reported to show biases away from fixation in visual bisection tasks, which may share the same origin as visual PSA biases.⁴⁰ Biases in learning spatial layout or locations of landmarks using residual vision have also been reported in patients with peripheral field loss.⁴¹ Further, there is a rich body of literature in normal vision focusing on the associations between visual spatial perception and visually guided action,⁴² which is indicative of possible associations between visual PSA and heading errors in low vision during navigation. Issues with auditory PSA may further add safety risks, as affected individuals may experience disorientations to sound and fail to respond to hazards in unfamiliar settings.

We acknowledge several limitations of the current study. Our study focused on measuring PSA and has not yet established the relationship between PSA and natural tasks in more complex visual and acoustic environments. Our horizontal field test was relatively coarse and could not measure visual field beyond 100°. Using a more established testing instrument such as the Goldmann visual field test could provide a more precise measurement of the binocular visual field. However, we note the merit of our customized test in that it reflects natural binocular viewing and takes a shorter time to complete (less than 5 minutes). This test was sufficient in revealing the reduced visual field and increased lateral binocular field asymmetries in our participants with vision impairment. Further, our study had a limited sample size to provide detailed comparisons between low-vision participants with central and peripheral field loss. Finally, it would have been informative to include a group of participants with normal vision and hearing impairment to understand the impacts of hearing impairment on auditory PSA.

To summarize, we reported the impact of vision impairment on the bias and variability in PSA using residual vision, the cross-modal association between visual PSA and auditory PSA, and the intact proprioceptive PSA in our low-vision participants. The lack of correlation between vision asymmetries and the laterality of visual PSA bias or between hearing asymmetries and the laterality of auditory PSA bias highlights the importance of directly measuring visual and auditory PSA in vision rehabilitation related to navigation and spatial orientation. Important follow-up studies should include investigations into the relationships between visual, auditory, and proprioceptive PSA and real-life spatial tasks, as well as rehab methods for improving the accuracy and precision in these tasks for individuals with low vision.