Fifty-eight participants were recruited from the Minnesota Laboratory for Low Vision Research, and the Envision Low Vision Rehabilitation Center. Inclusion criteria were age 18 years or older, self-reported total or near-total blindness with no form vision (i.e., no light perception, light perception, hand motion, or finger counting), and no cognitive impairments. This study was approved by the University of Minnesota Institutional Review Board (IRB number: STUDY00001360) and followed the Declaration of Helsinki (World Medical Association, [24]). Verbal consent was approved by the Institutional Review Board as part of this survey protocol. Consenting information was incorporated as the first section of the survey and read to the participants by the researcher. Participants can only proceed to the main survey after providing their verbal consent. The survey data was collected from 02/21/2022–07/14/2022.

Based on the level of self-reported residual vision, 34 participants were totally blind without light perception, and 24 participants were near-total blind with light perception, finger counting or hand motion. Out of all the 58 participants, 28 self-reported normal hearing (Blind-NH), and 30 participants self-reported with hearing loss. Because of the broad range of severity and onsets of hearing loss in our participants, we will use the term “hard of hearing” (Blind-HH) when referring to this participant group, per National Association of the Deaf recommendations [25].

The DS-SLQ includes descriptions of 35 everyday tasks that require localization of common objects around them like people or cars. For each task, participants responded to how difficult it was to complete each task with their vision alone or hearing alone based on their everyday experiences, using ratings from 1–5 with 1 being very easy and 5 being very difficult. They could respond “Not applicable” if unfamiliar with the tasks or situation posed in the question. The DS-SLQ was implemented as an online survey tool on Qualtrics and delivered through zoom videoconferencing or a phone call according to participant’s preference. More details of the development and validation of DS-SLQ can be found in Xiong et al. [23].

When administering the DS-SLQ to blind participants in the current study, vision ratings for each task were obtained for those with residual vision (e.g., light perception, hand motion, or finger counting), while were omitted for individuals without light perception. Responses were registered on Qualtrics by the researcher.

Prior to completing the DS-SLQ, participants first completed screening questions asking about their demographical information, including sex at birth and age, onsets for vision and/or hearing loss, rehabilitation history such as Orientation & Mobility training, and binaural asymmetry and use of hearing aids for those with hearing loss. Further, participants were asked to provide their history of using echolocation. Echolocation was defined for participants as: “The ability to sense your environment and objects surrounding you by listening to the echoes returned when making a sound such as mouth clicks, finger taps, cane taps, etc.”. Participants who self-reported as echolocators were further asked to rate their expertise in echolocation from 1–5, with the rating 1 being when they were able to echolocate but not very well, and rating 5 being an expert at echolocating. The entire survey took about 50 minutes to 1.5 hours in total.

All data analysis was performed using R statistical software (R Core Team [26]). Responses from the DS-SLQ were analyzed using the method of successive dichotomizations (MSD), using the R package ‘msd’ (Bradley & Massof [27]). MSD yields two primary outcomes: the item measure, indicating the difficulty of each task (higher values reflect more difficult tasks), and the person measure, representing an individual’s perceived spatial ability across all tasks (higher values indicate better perceived ability). A key feature of the DS-SLQ is that each task includes paired vision and hearing ratings, allowing assessment of spatial localization using either sense. To directly compare the difficulty of completing each task via vision versus hearing, a combined MSD analysis was conducted to align vision and hearing item difficulties on the same scale. The calibrated item measures were then anchored (fixed) in two separate MSD analyses, one for vision-only ratings and one for hearing-only ratings, to estimate perceived spatial abilities for each modality independently (see Xiong et al. [23]).

To evaluate the psychometric properties of the DS-SLQ for blind individuals, we examined two key psychometric properties: unidimensionality and targeting. Unidimensionality refers to a single underlying attribute (e.g., perceived auditory spatial ability in this case) that is being measured by the items of an instrument. We examined the information-weighted mean square fit (infit) statistics — acceptable infit values were within a range of 0.5–1.5, in accordance with COSMIN recommendations [28] — and conducted a principal component analysis to estimate the variance explained by the first principal component. Targeting refers to the degree to which the distribution of items measures matches that of the person measures. Poorly targeted instruments may show ceiling or floor effects, leading to highly imprecise estimates of person measures. Targeting was assessed for the DS-SLQ by comparing the distributions of item measures and person measures, to examine whether the perceived difficulty of the spatial localization tasks match the perceived spatial localization abilities in our sample of blind individuals.

Descriptive statistics summarized key characteristics of participants, including age, onset of vision and hearing loss, use of hearing aids, hearing asymmetry, echolocation skills, and Orientation & Mobility training history. Multivariable regression models were constructed to explore factors associated with perceived sound localization abilities (as reflected by DS-SLQ hearing person measures). An ANOVA was used to obtain the significance of each factor. Effect size was estimated using η² or partial η².

Table 1 summarizes the characteristics of participants in the Blind_NH and Blind_HH groups. The Blind_HH group was slightly older than the Blind_NH group (51.6 vs 59.5 yrs, F(1,55) = 4.46, p = 0.039, η² = 0.08). Most of the participants had early onset vision loss, with an average onset of 8 yrs for the Blind_NH group and 10.5 yrs for the Blind_HH group. The primary causes for the blindness include retinitis pigmentosa and retinopathy or prematurity. Most of the participants in the Blind_HH group had later onset of hearing loss than their vision loss (mean difference = 26.4 yrs, F(1, 56) = 22.17, p < 0.001, η² = 0.28). Only four of the Blind_HH participants reported hearing loss from birth due to the same medical condition as their vision loss (3 with Usher Syndrome, 1 with Marshall Syndrome).

In total, 96.4% and 93.3% of the participants in the Blind_NH and Blind_HH groups have received O&M training, respectively. In the Blind_HH group, 70% of the participants reported one ear to be better than the other, and 70% reported as hearing aids users.

Figs 1A illustrates the distributions of item and person infits for blind participants in our sample. Approximately 84% of the item measures and 78% of the person measures fall within the desired infit range of 0.5–1.5, indicating acceptable unidimensionality of the DS-SLQ in assessing perceived sound localization abilities among blind individuals. Principal component analysis on the raw ratings further confirmed the unidimensionality, with the first principal component explaining 66.5% of the variances.

Fig 1B illustrates “targeting” of DS-SLQ for blind participants by showing the distributions of the person measures (upper panel) and item measures (lower panel). Item measures ranged from −3.6 to 3.1 logits, while the person measure ranged from −4.0 to 2.9 logits. When comparing these distributions, 98% of the person measures landed within the range of item measures, demonstrating excellent targeting of DS-SLQ for blind individuals.

To understand factors associated with perceived sound localization abilities, in a multivariable regression model across all participants we treated hearing person measures as the dependent variable, age (in years), onset of blindness (in years), residual vision status (residual vision vs. no light perception), and hearing status (NH vs. HH) as independent variables. Although we were also interested in O&M training, the very high prevalence of O&M training history in both groups prevented us from including this factor.

The result showed that later onset of blindness (β = −0.04, F (1, 52) = 8.03, p = 0.007, partial η² = 0.13) and presence of self-reported hearing loss (F (1, 52) = 7.48, p = 0.009, partial η² = 0.13) were associated with significantly lower perceived sound localization abilities. Because the onsets of blindness clustered around 0 years old, we repeated this analysis on those who acquired blindness after birth and found the same effect of the onsets (β = −0.04, F (1, 26) = 6.18, p = 0.020, partial η² = 0.19). There was no interaction between the onset of blindness and hearing status, suggesting that the effects of onsets of blindness were uniform across the blind individuals. Figs 2A and 2B visualized the distribution of sound localization abilities across these two factors, respectively.

Next, we focused on individuals with hearing loss, asking whether the onsets of hearing loss and use of hearing aids played additional role in perceived hearing spatial abilities. In a multivariate regression model on the Blind_HH group, we again treated hearing person measures as the dependent variable, age, onset of blindness, residual vision status, as well as onset of hearing loss (in years), perceived binaural asymmetry (yes or no), and use of hearing aids (yes or no) as independent variables. The result again demonstrated a significant effect of the onset of blindness (β = −0.04, F (1, 20) = 6.99, p = 0.016, partial η² = 0.26); in addition, earlier onset of hearing loss was associated with poorer perceived sound localization abilities (β = 0.04, F (1, 20) = 12.16, p = 0.002, partial η² = 0.38) (Fig 2C). Perceived binaural asymmetry or use of hearing aids was not a significant factor.

We considered echolocation as a separate factor as it is a form of localization using self-generated sounds, different from locating external sounds that DS-SLQ intends to measure. In total, 66% of the blind participants self-reported to be echolocators. Participants with later onset of blindness (β = −0.07, χ² (1) = 9.27, p = 0.002) and presence of self-reported hearing loss (χ² (1) = 6.71, p = 0.010) were found to be less likely to be echolocators. It is interesting to note the similar effects of these two factors on DS-SLQ measured sound localization as mentioned above. In Fig 3A, we summarized the percentages of echolocators in the Blind_NH and Blind_HH groups, stratified by whether the onset of blindness was at birth.

Among participants who self-reported as echolocators, we further explored the associations between self-rated echolocation expertise and the perceived sound localization abilities measured by DS-SLQ. As shown in Fig 3B, there was a significant correlation between these two measures (r = 0.58, p < 0.001), independent of hearing status.

Amongst the 24 participants with residual vision who self-reported light perception, hand motion, or finger counting, 16 gave valid vision ratings (9 from Blind_NH and 7 from Blind_HH groups, respectively) on at least 2 of DS-SLQ tasks. Vision person measures were thus obtained from these 16 participants, allowing a more detailed description of the perceived utility of their residual vision in visual localization. We conducted the following exploratory analysis, due to the small sample of available vision measures.

Fig 4 illustrates the comparison of perceived vision and hearing localization abilities for these 16 individuals, with dark orange representing the Blind_NH group and dark cyan presenting the Blind_HH group. There was no significant correlation between the vision and hearing abilities in either group. When comparing the two localization abilities, more participants in the Blind_NH had higher hearing ability than vision, while more participants in the Blind_HH had higher vision abilities than hearing, suggesting that even very limited vision by clinical standard can play a critical role in blind individuals, especially in the presence of co-occurring hearing loss.