A blind human expert echolocator shows size constancy for objects perceived by echoes

Some blind humans make clicking noises with their mouth and use the reflected echoes to perceive objects and surfaces. This technique can operate as a crude substitute for vision, allowing human echolocators to perceive silent, distal objects. Here, we tested if echolocation would, like vision, show size constancy. To investigate this, we asked a blind expert echolocator (EE) to echolocate objects of different physical sizes presented at different distances. The EE consistently identified the true physical size of the objects independent of distance. In contrast, blind and blindfolded sighted controls did not show size constancy, even when encouraged to use mouth clicks, claps, or other signals. These findings suggest that size constancy is not a purely visual phenomenon, but that it can operate via an auditory-based substitute for vision, such as human echolocation.

A number of studies have confirmed the ability of blind human echolocators to extract information from echoes about object features such as shape, location, motion, and material (for reviews see Kolarik, Cirstea, Pardhan, & Moore, 2014;Stroffregen & Pittenger, 1995). Furthermore, echoes are informative for determining the size and distance of objects Rice, Feinstein, & Schusterman, 1965;Teng & Whitney, 2011). Although it has been confirmed that expert echolocators (EE) can extract size and distance information from echoes in isolation, there has been no direct investigation on the echolocation of these factors in combination.
Integrating size and distance cues is critical for accurate perception of object size. For example, the visual system must overcome the fact that altering the distance of an object from the eyes results in changes to the retinal image size. Despite the differences in retinal image size, however, people's percept reflects the true physical size of the object. This perceptual phenomenon has been termed "size constancy" (Holway & Boring, 1941).
In echolocation, a similar situation may arise. Specifically, the size of the acoustic angle of a soundreflecting surface decreases as its distance increases. Importantly, changes in the size of a sound-reflecting surface (regardless of changes in distance) lead to changes in both the level and spectrum of the reflected sound (e.g. Heinrich, Warmbold, Hoffmann, Firzlaff, & Wiegrebe, 2011). At the same time, the distance of a sound-reflecting surface is reliably coded via pulse-echo delays. Hence, the acoustic information present in echolocation is sufficient for accurate computation of the physical size of a sound-reflecting surface. Given that the human brain shows size constancy in at least one modality (i.e. vision), and given that echolocation in principle provides the information necessary to achieve size constancy, we were interested if size constancy is supported for objects perceived via reflected echoes.
To test this, we recruited a blind human EE who lost his sight in early infancy due to idiopathic optic nerve atrophy and has no residual vision. EE developed echolocation techniques on his own during early childhood, and, now in his late 50s, he continues to use echolocation on a daily basis.

Methods Participants
As mentioned earlier, we recruited one blind echolocating participant (male, aged 57 years). Hearing tests revealed that EE's pure tone thresholds were within the normal range up to 4 kHz, but that he has mild hearing loss beyond 4 kHz, consistent with his age. As control groups, we recruited 10 blind (5 female; mean age = 35.9 years) and 10 sighted (5 female; mean age = 40.4 years) participants. Control participants reported normal hearing and no prior experience with echolocation. All participants were blindfolded and took part in three experiments at The University of Western Ontario (London, Canada). Testing took place in an echo-dampened room (2.75 × 3 m 2 , walls covered in 3.8-cm convoluted foam sheets). Testing procedures were approved by the University ethics board and participants provided informed consent. *Corresponding author. Email: mgoodale@uwo.ca Neurocase, 2015Vol. 21, No. 4, 465-470, http://dx.doi.org/10.1080/13554794.2014 Stimuli and general procedure Objects used in the experiments included three physical sizes of circles and horizontally oriented rectangles that could be placed at one of three distances (Figure 1). The shapes were designed so that the medium-sized shape was twice the size of the small shape, and the large-sized shape was three-times the size of the small shape. Furthermore, each size of circle and rectangle were near-equated in total area. In accordance with the proportional size relationship between the objects, the distances were chosen so that the middle location was twice the distance of the near location, and the far location was three times the distance of the near location. Although shape is not a relevant cue for size constancy, we included the different shapes to increase the attentional demands of the tasks. The objects and distances used depended on the particular experiment being run (see later and Figure 1). Before beginning the experiments, all participants were allowed to haptically explore the objects.
Objects were positioned in isolation directly ahead of the participant, with the center at ear level. Between trials, participants wore ear-bud headphones playing white noise to mask any noise cues related to changing the stimulus display. Once the trial began, participants were given a maximum of 20 s to scan the object using the echolocation technique of their choice.
Importantly, blind and sighted controls received no echolocation training prior to participation. The purpose of the control participants in this case was to control for performance that could be attributed to factors other than echolocation expertise (e.g. neuroplasticity due to blindness, ambient sounds, and sounds from the movements of the experimenter). The mouth-click, finger-snaps, and other signals were explained to control participants and they were free to use the technique of their choosing.

Experiment 1: distance
To investigate distance perception, we used three circular objects and presented them at each of the three distances ( Figure 1(b)). Participants provided a verbal response indicating the perceived distance ("near", "middle", "far"). There were a total of 36 trials (four repetitions per object, per distance).

Experiment 2: acoustic size
To investigate the detection of size, we utilized all six objects but presented them only at the middle distance. The participants' task was to determine both the size and shape of the object (e.g. "small circle", "medium rectangle"). Since distance was kept constant, identification of size in these cases was equivalent to identification of acoustic size (i.e. area in degrees of acoustic angle), but it did not require size constancy. There were a total of 36 trials (six repetitions per object). The possible objects to be used in each of the experiments included three physical sizes of circles and rectangles (the medium-sized objects were double the size of the smallest objects, and the large objects were triple the size of the smallest objects). The objects were created from 0.5-cm-thick foam board and covered with aluminum foil and were positioned on a 0.6-cm-diameter pole. (b) The basic layout is shown for each individual experiment. The top panel shows the distance(s) at which the objects could be placed (specific objects used in each experiment are shown in the bottom panel). Note that the spacing of the distances is proportional to the sizes of the objects (i.e. the middle position is twice the distance of the near position, and the far position is three times the distance of the near position). For all tasks, participants stood and were permitted to move side-to-side as well as up and down to allow them to scan the object (for a maximum of 20 s) but were not permitted to move toward or away from the object. Note that for Experiment 3, only the near and far positions and small and large objects were used. This allowed for a direct test of size constancy.

Experiment 3: size constancy
To examine size constancy, i.e. correct perception of physical size, we designed the experiment in accordance with the principles of visual size constancy. That is, we used only the small and large objects and placed them only at the near and far distances. The small objects at the near location would have the same acoustic size (i.e. same echo intensity and aperture) as the large objects at the far position. The participants indicated the size and shape of the object (e.g. "small rectangle", "large circle"). There were a total of 40 trials (five repetitions per object, per distance).

General details
Our initial analyses were aimed at determining if there were any differences in performance between blind and sighted control groups for each of the three experiments. We ran mixed analysis of variance tests, with the experimental manipulations as the repeated variable, and "group" as the between subjects variable for each experiment. The analyses revealed no significant effect of "group" for any of the three experiments ( Figure 2). Therefore, for all subsequent analyses, we combined blind and sighted control groups into a single control group.
To compare EE's performance to the combined control group, we ran two types of analyses. First, we used t-tests and effect size measures modified to compare a single case to a control group (Crawford, Garthwaite, & Porter, 2010;Crawford & Howell, 1998). The analyses were conducted on overall percentage correct performance in each experiment. We supplemented these analyses by comparing both EE's and the control group's performance to chance for each experiment using binomial tests and traditional t-tests, respectively. Instead of overall performance, these analyses included the separate conditions within each experiment. An initial analysis had revealed no differences between circular and rectangular shapes for Experiments 2 and 3. Thus, for subsequent analyses, performance was collapsed across shape.

Experiment 3: size constancy
As mentioned earlier, Experiment 3 was aimed at determining if an EE can integrate acoustic size and distance information to determine the true physical size of objects. The modified t-test analysis revealed that EE performed well above the level of the control group (Figure 3(a) and Table 1). In fact, EE's best performance across all three experiments was in Experiment 3. Comparisons against chance revealed that EE performed significantly better than chance level (25%) in three of the conditions (small- Figure 2. Average percent correct performance (±SEM) of the blind non-echolocating and blindfolded sighted control groups on the distance, acoustic size, and size constancy experiments. Mixed analysis of variance tests for each of the three experiments revealed no significant main effects or interactions for any of the three experiments (similarities in performance are easily seen in the figure). Therefore, for all subsequent analyses, the control groups were combined to form a single control group. near: p < .0005; large-near: p < .0005; large-far: p < .0005), but his performance failed to reach significance for the "small-far" condition (p = .078). It is important to note, though, that EE's performance in this condition was largely driven by errors in shape and not in size (see Figure 4(b)). The control group did not show above chance performance for any of the conditions (small-near: t(19) = .667, p = .513; small-far: t(19) = .17, p = .867; large-near: t(19) = 1.082, p = .293; large-far: t(19) = 1.255, p = .225).
Overall, the results of Experiments 1 and 2 agree with previous reports in the literature on size and distance perception in human echolocation . Most importantly, Experiment 3 goes beyond this and shows that a blind echolocator can accurately identify the size of an object independent of its distance. These results suggest that size constancy operates for object size perception via echolocation.

Discussion
The aim of the current experiment was to determine if size constancya perceptual phenomenon linked to visionalso operates for echolocation. We showed for the first time that a blind EE could consistently and reliably discriminate between objects of different sizes presented at different distances. Furthermore, even when the large object was placed at the far distancewhich would "equate" its acoustic size to the small object at the near distancethe EE had no trouble accurately identifying the object. Blind non-echolocators and blindfolded sighted controls did not show size constancy, suggesting that echolocation expertise was responsible for EE's performance, rather than neuroplasticity due to blindness or ambient sounds.
The current findings support the use of echolocation as a viable resource for the blind, because the ability to accurately determine the physical size of objects has Figure 3. Results of the individual case analyses for the distance, acoustic size, and size constancy experiments. (a) The results of the modified t-tests comparing the expert echolocator's overall performance (as percent correct) to the combined control group's performance are shown. Significant differences are indicated by asterisks. The Bayesian effect sizes (with error bars showing 95% confidence intervals (CIs)) of each individual t-test are shown in (b). The effect size was calculated using adapted z scores (Crawford et al., 2010). (c) The "abnormality" of the case's scores is presented, which shows the percentage of the control population (with 95% CIs) that would obtain a lower score than the case. immediate benefits for navigation. In addition, the current findings broaden our basic understanding of the technique. For example, previous studies using functional magnetic resonance imaging have implicated "visual" brain areas for echolocation (Arnott, Thaler, Milne, Kish, & Goodale, 2013;Thaler, Arnott, & Goodale, 2011;Thaler, Milne, Arnott, Kish, & Goodale, 2014). The current findings suggest further parallels between vision and echolocation, in that both modalities show size constancy. This suggests that similarities in brain activity may also signify similarities in terms of behavioral principles, and future research should address these possibilities.
As laid out in the discussion at the start of the article, object size information can be inferred from the overall intensity of the echo as well as spectral changes caused by the "spread" of angles from which the echoes arrive at the ears (aperture). The contribution of each of the cues may depend on the size of the object, with evidence suggesting that overall intensity cues are best suited for smaller objects whereas aperture cues are most relevant for larger objects, at least in the case of echolocating bats (Heinrich et al., 2011). Furthermore, there are binaural cues to size (Holderied & von Helversen, 2006). In terms of object distance, the cue that indicates distance most reliably is the time delay between the outgoing signal and the returning echo, and this cue is independent from other aspects of the sound. Thus, echolocation has information sufficient for size constancy.
Although the current study is the first investigation of size constancy in human echolocation, Heinrich and Wiegrebe (2013) tackled the question in the context of bat echolocation. Interestingly, their results suggested that bats do not show size constancy, at least for the perception of virtual objects. This is curious, because echolocating bats do encode both object aperture (i.e. acoustic size) (Heinrich et al., 2011) and pulse-echo delays (i.e. distance) (Wenstrup & Portfors, 2011). Therefore, it is surprising that bats would not show size constancy considering the clear neural representation of both size and distance. Contrary to these unresolved findings in bats, our results clearly show that a human echolocator has stable, absolute size perception. Considering that human echolocators may lack the sophisticated neural mechanisms that have evolved in bats, our findings, then, warrant further investigation into potential size constancy mechanisms in bats and other echolocating species. Figure 4. Results from tests against chance for the expert echolocator (binomial tests) and combined control group (t-tests). Participants' performance (shown as percent correct) was collapsed across shape for Experiments 2 and 3. Chance performance for each of the tasks is indicated by the dashed lines and significant results are indicated by asterisks. The distribution of EE's responses (b) is shown for the size constancy task. Although EE performed statistically at chance for the "small-far" condition, the error distributions show that this decrease in performance was driven by errors in shape judgment and not size judgment, thus supporting size constancy in these cases.