Perception of combined translation and rotation in the horizontal plane in humans

During human ambulation, rotation and translation occur simultaneously, such that lateral translation is coupled with inward rotation. However, there has been little previous work on how these combined movements are perceived. The current study determined the bias and threshold of human perception during such combined movements and found concurrent rotation biases perception of lateral translation, such that translation occurring during ambulation is likely to be perceived only as rotation.

Abstract

Thresholds and biases of human motion perception were determined for yaw rotation and sway (left-right) and surge (fore-aft) translation, independently and in combination. Stimuli were 1 Hz sinusoid in acceleration with a peak velocity of 14°/s or cm/s. Test stimuli were adjusted based on prior responses, whereas the distracting stimulus was constant. Seventeen human subjects between the ages of 20 and 83 completed the experiments and were divided into 2 groups: younger and older than 50. Both sway and surge translation thresholds significantly increased when combined with yaw rotation. Rotation thresholds were not significantly increased by the presence of translation. The presence of a yaw distractor significantly biased perception of sway translation, such that during 14°/s leftward rotation, the point of subjective equality (PSE) occurred with sway of 3.2 ± 0.7 (mean ± SE) cm/s to the right. Likewise, during 14°/s rightward motion, the PSE was with sway of 2.9 ± 0.7 cm/s to the left. A sway distractor did not bias rotation perception. When subjects were asked to report the direction of translation while varying the axis of yaw rotation, the PSE at which translation was equally likely to be perceived in either direction was 29 ± 11 cm anterior to the midline. These results demonstrated that rotation biased translation perception, such that it is minimized when rotating about an axis anterior to the head. Since the combination of translation and rotation during ambulation is consistent with an axis anterior to the head, this may reflect a mechanism by which movements outside the pattern that occurs during ambulation are perceived.

NEW & NOTEWORTHY

During human ambulation, rotation and translation occur simultaneously, such that lateral translation is coupled with inward rotation. However, there has been little previous work on how these combined movements are perceived. The current study determined the bias and threshold of human perception during such combined movements and found concurrent rotation biases perception of lateral translation, such that translation occurring during ambulation is likely to be perceived only as rotation.

human motion perception is an important issue that in the past, has largely been examined with perception of isolated translation or rotation (Benson et al. 1986, 1989; Grabherr et al. 2008; Mallery et al. 2010; Roditi and Crane 2012a; Soyka et al. 2011; Valko et al. 2012; Walsh 1961). Translation and rotation are sensed by separate receptors: the otoliths (translation) and semicircular canals (rotation). These two signals converge as early as the vestibular nuclei of the brain stem (Bush et al. 1993; Dickman and Angelaki 2002; Jian et al. 2002; Uchino et al. 2005; Zhang et al. 2001; Zhou et al. 2006). It makes sense that this convergence should occur; during many common activities, such as ambulation, canal and otolith stimulation occurs simultaneously (Grossman et al. 1989). Thus the knowledge of vestibular perception, when both semicircular canal and otolith signals are present, is important for understanding real-world situations, such as ambulation.

Most of the previous work on effects of canal-otolith interaction in the vestibular system has been based on resolving the tilt-translation ambiguity (Angelaki and Cullen 2008; Angelaki et al. 2004; Green and Angelaki 2007; Merfeld et al. 1999; Zhou et al. 2006). There is also evidence that canal and otolith signals are combined at the earliest stages of vestibular processing (Dickman and Angelaki 2002; Shaikh et al. 2005; Zhou et al. 2006), suggesting a mechanism by which this could occur. However, tilt is only one situation in which linear acceleration is combined with rotation. During human ambulation, head rotation is combined with translation, such that lateral translation is associated with an inward rotation (Crane and Demer 1997; Grossman et al. 1989), a combination of translation and rotation that also occurs with rotation about an axis anterior to the head. In this situation, the orientation of the gravity vector relative to the subject remains constant. Although this does not completely avoid the possibility of the tilt-translation ambiguity, it opens the door to examining other types of canal-otolith interaction for movements that are limited to the horizontal plane.

The relative contributions of combined otolith and semicircular canal stimulation in the perception of movement in the horizontal plane have been examined. In one experiment, subjects underwent a series of movements in the horizontal plane, after which subjects oriented a pointer toward a previously seen target or drew their perceived trajectory (Ivanenko et al. 1997). These experiments suggested that the semicircular canals dominated the perceived trajectory or, at least, biased translation perception, although thresholds were not measured. In another study, patients experienced rotation about a head-centered axis and about an eccentric axis, which included otolith stimulation (Israel et al. 2005). They were later asked to reproduce the rotation using a joystick. It was found that the additional otolith information with the eccentric axis rotation made it more difficult for subjects to reproduce the rotation angle, suggesting a dominate role for the canals, although they did a better job at reproducing the velocity profile of the stimulus when otolith information was present.

With the use of more quantitative methods, a later series of experiments measured thresholds by having subjects report the presence or absence of angular rotation in the presence of forward-backward translation or vice versa (MacNeilage et al. 2010b). In those experiments, yaw rotation-detection thresholds were not influenced by the presence of translation, but translation detection was greatly increased when concurrent rotation was present. In another study, the thresholds of right-left discrimination were measured using synergistic rotation and translation (Soyka et al. 2015). They considered two competing models—one in which either rotation or translation was used and another in which information from both organs was optimally integrated based on cue reliability. The results of Soyka et al. (2015) were most consistent with integration of signals from both organs, as the thresholds were lower than would be predicted from either rotation or translation alone.

The previous studies of combined translation-rotation perception in the horizontal plane do not paint a clear picture. Studies that focused on stimulus magnitude found that rotation dominated perception (Israel et al. 2005; Ivanenko et al. 1997), although it is difficult to know the relative contribution of translation alone, since neither study examined it independently. When thresholds were determined, there were conflicting results, with one study demonstrating higher thresholds for translation when rotation was present (MacNeilage et al. 2010b) and the other demonstrating lower thresholds for the combined stimuli (Soyka et al. 2015). The differences between these studies may be due to the relationship between rotation and translation, which always occurred in the same relative direction (i.e., the axis of rotation was always behind the head) in one study, so that either rotation or translation could be used to determine the direction of both (Soyka et al. 2015) but was directionally independent in the other study (MacNeilage et al. 2010b). During ambulation, translation and rotation tend to occur in a fixed relationship (Crane and Demer 1997), which raises the possibility that perception could assume such a relationship, although such an assumption would not hold true in every situation. The MacNeilage et al. (2010b) study varied whether rotation and translation were present or absent, which forced them to be considered independently; this may be the reason why they found thresholds were the same or higher when translation and rotation occurred simultaneously. However, there are also other differences between these studies, including the type of motion studied and task. Neither of these studies addressed the issue of bias or the potential for rotation to influence the perceived direction of translation and vice versa, because subjects did not report direction (MacNeilage et al. 2010b). Translation and rotation in the Soyka et al. (2015) study were always presented, such that either determined the direction of both.

The current study examines perception of concurrent rotation and translation within the horizontal plane. These types of motion often occur concurrently during common activities, such as ambulation, but it remains unclear how they are perceived in these situations. One hypothesis is that translation and rotation are perceived independently and linearly combined analogously to how translation and rotation interact in the vestibulo-ocular reflex (Crane and Demer 1999; Seidman et al. 2002). However, unlike eye movement, in which the vestibulo-ocular reflex needs to compensate for both translation and rotation, no such common output is needed for perception. As an alternative hypothesis, it is possible that during complicated combined motion, perception may be filtered so that the more salient cues that might indicate an unexpected situation are more likely to be perceived. Both thresholds and bias are measured during combined rotation and translation to examine these possibilities.

METHODS

Subjects.

A total of 23 subjects was enrolled; however, 6 could not report the direction of translation when rotation was present and were excluded. Seventeen human subjects (10 women, 7 men), who were between the ages of 20 and 83 (mean 39; SD 21) completed the protocol. For the purpose of analysis, subjects were divided into groups of younger controls (age under 50; range 20–45; mean 27; SD 7; 6 men; 6 women) and older controls (age 50 or greater; range 53–83; mean 69; SD 12; 4 women; 1 man). The 50-yr cutoff was used, because it was previously found to be an appropriate cutoff for vestibular function (Crane 2014; Roditi and Crane 2012a). All subjects were naïve to the design of the experiment, and all subjects were healthy and had undergone screening for history of dizziness, vertigo, hearing, neurological, and vision issues. Subjects who reported a migraine history were excluded. The study was approved by the University of Rochester Research Science Review Board. All subjects granted written, informed consent before participation, and subjects were paid for their participation.

Equipment.

During the study, motion was provided with a six degrees-of-freedom Hexapod motion platform (model 6DOF2000E; Moog, East Aurora, NY). The setup has previously been described in the current laboratory (Roditi and Crane 2012a). Subjects sat upright in a padded racing seat with lumbar and seat bolsters mounted to the platform (model FX-1; Corbeau, Sandy, UT). Subjects wore an appropriately sized helmet that was fixed securely to the Hexapod motion platform using a custom-built structure, ensuring that head motion was closely coupled to the platform. Subjects were carefully positioned, such that the center of the platform was directly below the midline between the subject's external auditory canals. Platform noise was masked with white noise produced from two platform-mounted speakers. The intensity of the white noise was varied during platform motion so that peak masking occurred at the time of peak motor noise. The intensity of the noise was independent of the direction of platform motion.

The platform motion was a sinusoid in acceleration with a duration of 1 s (1.0 Hz). Stimuli were designed with a maximum displacement of 7° (yaw) or 7 cm in the horizontal plane. Both forward-backward translation (surge) and left-right translation (sway) are considered. Thus peak velocity was 14°/s for rotation and 14 cm/s for both types of translation, respectively. This motion profile was used because it contained no discontinuities in acceleration, velocity, or position. It has previously been used in the current laboratory (Crane 2012a; Roditi and Crane 2012a), as well as by others (Benson et al. 1986; Grabherr et al. 2008; MacNeilage et al. 2010b; Valko et al. 2012).

Immediately after the stimulus presentation, there was an audible tone (2, 500 Hz; 0.125 s in rapid succession) that cued the subjects to report the direction of motion they perceived by pressing a button. After the response was reported, the platform returned to the starting position. After returning, a tone was presented to indicate the next stimulus was ready, and the subjects pressed the center button to deliver the next stimulus.

Experimental procedure.

The experiments were organized into several blocks of trials. The order in which these blocks were delivered was randomized for each participant. Control experiments were done in three blocks with a single stimulus presented: surge (fore-aft motion), sway (left-right motion), or yaw (rotation about an earth vertical axis). During each of these blocks, there were two independent staircases that were randomly interleaved: one staircase began with the largest (7 cm or 7°) leftward stimulus, and the other began with the largest rightward stimulus. Subjects were always able to identify reliably this largest stimulus correctly. After each response, the subsequent stimulus was shifted in the direction opposite the response. Thus after a 7-cm rightward stimulus, which was identified as right, the next stimulus was delivered, 1.6 cm more leftward (i.e., 5.4 cm to the right). The initial step size was 1.6 cm or 1.6°. Within each staircase, the step size was decreased by one-half when the direction of responses reversed to a minimum of 0.1 cm or 0.1°. The validity of these methods is well established (Green 1993; Leek 2001; Levitt 1971; Treutwein 1995). Furthermore, if the response was in the same direction three times in a row, then the step size doubled to a maximum of 1.6 cm or 1.6°. Each staircase could move through zero so that later in the staircase, stimuli could be delivered in either direction. This method tended to focus stimuli near the point of subjective equality (PSE) but also delivered enough stimuli away from the PSE that threshold could be determined. It was thought to be a reasonable method of the current study, because both the PSE (i.e., mean of the psychometric function or bias) and threshold (i.e., width of the psychometric function) were of interest. Each staircase included 25 stimulus presentations (50 total within the trial block).

The influence of multiple stimuli was assessed in blocks of trials with combined translation and rotation (Fig. 1). In these experiments, there was a fixed, distracting stimulus delivered simultaneously with the test stimulus, although the subject was asked to report only the test stimulus direction. Blocks of trials included randomly interleaved sets of staircases with distracting stimuli in opposite directions. There were four blocks of trials of this type that are named as test distractor: surge-yaw, sway-yaw, yaw-surge, and yaw-sway. The distracting stimulus was always 7 cm or 7° (14 cm/s or 7°/s peak velocity). Each distracting stimulus had two staircases that started with test stimuli in opposite directions. Thus each block of trials, which included a distracting stimulus, had a total of 4 randomly interleaved staircases for a total of 100 stimulus presentations (Fig. 2). It is recognized that more stimulus presentations may have allowed the parameters of the psychometric function to be fit with narrower confidence intervals, but this length was chosen to improve subject alertness by keeping trial blocks to an appropriate length. This is in the same range of trials used in other human vestibular perception studies. Although some studies have used 150 stimulus presentations per trial block (Soyka et al. 2015) or 110 per trial block (Roditi and Crane 2012a), others have used 100, as done in the current series (Yi and Merfeld 2016). Many other trials have had fewer: 70–80 (Valko et al. 2012), 50 (Benson et al. 1989), or <50, on average (Grabherr et al. 2008).

Fig. 1.

Trial blocks, including combined rotation and translation. Each condition included both rotation and translation with subjects asked to judge left-right sway (top), forward-back surge (middle), or left-right rotation (bottom).

Fig. 2.

Sample date from an individual subject (#5, a 23-yr-old woman) for the sway-yaw condition. During the stimulus presentations, there was a 7° (14°/s peak velocity) leftward rotation (A and B) and similar rightward rotation (C and D). The subject was asked to report the amount of left-right sway translation. A and C: the circles represent this subject's responses and are drawn in proportion to the number of stimuli presented for a given combination. The smaller circles represent a single stimulus presentation, whereas the larger circles represent 4 stimuli. The solid lines represent the best fit of a cumulative Gaussian function to the responses. The values in parentheses represent the 95% confidence interval (CI). The same data are shown in the time domain (B and D). Four randomly interleaved staircases are shown. Each panel includes 2 staircases: 1 that began with a 14-cm/s translation (light) and another that began with a −14 cm/s translation (dark). The direction of each response is shown by the orientation of the triangles.

The influence of combined rotation and translation was also examined in a block of trials in which the axis of rotation was varied (sway-axis and surge-axis). These trials were, in some ways, similar to the surge-yaw and sway-yaw condition in that both translation and rotation occurred simultaneously, and subjects were asked to report the direction of translation. However, they differed from the surge-yaw and sway-yaw conditions in that the translation occurred about a curved trajectory. Attempts were also made to correlate these results with the surge-yaw and sway-yaw condition. In these trials, the distracting stimulus was ±7° (14°/s peak velocity) yaw rotation. In the sway-axis trial block, the axis was varied in the fore-aft direction (Fig. 1), and subjects were asked to report the perceived direction of left-right translation (sway). In the surge-axis block of trials, the axis was varied in the left-right direction (Fig. 1), and subjects reported if they perceived they were moving forward or backward (surge). Similar to the combined translation-rotation trials, there were independent pairs of staircases that started with the rotation axis located 100 cm away from center in either direction. The initial step size was 16 cm, which could be decreased to 1 cm with reverses in the perceived direction.

Data analysis.

Subjects' responses were fitted to a cumulative Gaussian function with a Monte Carlo method used to measure the mean (PSE), sigma (threshold), and confidence intervals, as described previously (Wichmann and Hill 2001a, b) and used in the current laboratory (Crane 2012b; Roditi and Crane 2012b), as well as by others (Fetsch et al. 2009; MacNeilage et al. 2010a). The curve fitting allowed lambda and gamma to be up to 0.05 to allow for a small lapse rate. The data from each subject were resampled 2,000 times to allow for multiple estimates of the mean and threshold with 95% confidence intervals (Fig. 2). This also allowed the level of significance between the two distributions to be determined, as previously described (Crane 2012b). The repeated-measures ANOVA was used to compare the bias between subjects and test conditions. Paired, two-tailed t-tests were used to compare data across subjects. Fishers exact test was used to determine if there was a significant difference in ability to do a task.

RESULTS

Seventeen subjects completed all of the test conditions. Six additional subjects began the protocol but were not able to complete it, because they were not able to do some of the tasks required. When a translation test stimulus was combined with a rotation distractor (e.g., sway-yaw, sway-axis), all six of these subjects were only able to report the direction of rotation and could not report translation, even at the maximum translation stimulus available with the current apparatus. All of these subjects reported that they could only perceive the rotation, were at the higher end of the age spectrum (47–86 yr, mean 68, SD 14), and could report translation when it was presented without rotation. When these subjects were considered by age, in those under 50, 12 of 13 were able to complete the task, and in those over 50, 5 of 11 were able to complete the task. With the use of Fishers exact test (two-tailed) this represented significantly fewer people over 50 who were able to do this task (P = 0.02). For the remainder of the analysis, only the 17 subjects who were able to complete all test conditions are reported.

The threshold of surge (fore-aft translation), in the absence of a distracting rotation, averaged 0.9 cm/s in subjects under 50 and was 2.4 cm/s in those older than 50 (Fig. 3A). However, there was a large amount of variation, especially in the older subjects, and the difference did not reach statistical significance (unpaired t-test, P = 0.08). When a distracting yaw stimulus was present, the thresholds increased significantly (paired t-test, P = 0.007). The threshold of sway (right-left translation) averaged 0.7 cm/s in those under 50 and was 2.9 cm/s over 50, although this difference also did not reach significance (unpaired t-test, P = 0.07). Similar to surge, there was a significant increase in the threshold when the translation was paired with yaw rotation (paired t-test, P = 0.001), although the increase was largely in the younger subjects (Fig. 3B).

Fig. 3.

Translation thresholds. Thresholds, as determined from the width (sigma) of the psychometric function, are shown for 17 subjects. The control condition in which only translation was presented (no yaw rotation) is shown with circles. Conditions in which there was a simultaneous distracting yaw motion are shown with triangles. Data for individual subjects (1–17) are in order of age; error bars represent the 95% CI. The CI tended to be larger with higher thresholds due to the fixed number of stimulus presentations in each trial block. Average data from subjects under age 50 (1–12) are shown in the column labeled Y (younger). Average data from the subjects 50 and over (13–17) are shown in the column labeled O (older). For the grouped data, error bars represent SD. A: surge (for-aft) translation. B: sway (lateral) translation.

When the threshold of yaw rotation was considered in the absence of a distractor, it was 0.9°/s in the younger (age <50) subjects and 1.6°/s in the older subjects. The difference with age was again not significant (P = 0.18). The thresholds were similar when a surge distractor was present at 0.9°/s in the younger group and 1.8°/s in the older subjects (Fig. 4A). The presence of a sway distractor raised the thresholds to 2.0°/s in younger subjects and 1.9°/s in older subjects (Fig. 4B), but this higher threshold with the distractor was not significant (P = 0.17, paired t-test).

Fig. 4.

Yaw rotation threshold. The control condition in which only rotation occurred (no translation) is represented by circles. Conditions in which there was a simultaneous distracting translation are shown with triangles. Data for individual subjects (1–17) are in order of age; error bars represent the 95% CI. The CI tended to be larger with higher thresholds due to the fixed number of stimulus presentations in each trial block. Average data from subjects under age 50 (1–12) are shown in the column labeled Y (younger). Average data from the subjects 50 and over (13–17) are shown in the column labeled O (older). For the grouped data, error bars represent SD. A: distracting stimulus was surge (for-aft) translation. B: distracting stimulus was sway (lateral) translation.

The mean or bias of the psychometric function was also considered. In cases where subjects reported the direction of surge alone or with a yaw distractor, there were occasional biases in individual test conditions, but on average, the 95% confidence interval of the bias contained zero (Fig. 5A). However, concurrent yaw rotation did tend to bias perception of sway direction in almost all subjects (Fig. 5B). The bias was similar in the older and younger groups, and such that in the presence of a 14°/s peak velocity leftward rotation, the PSE occurred with sway of 3.2 ± 0.7 (mean ± SE) cm/s to the right. Similarly, during 14°/s rightward motion, the PSE was with sway of 2.9 ± 0.7 cm/s to the left. The difference in the bias, based on the direction of rotation, was highly significant (paired t-test, P = 0.0001). Thus yaw rotation consistently biased the perception of sway (left-right translation) but not surge (forward-back translation).

Fig. 5.

Translation bias. The bias or mean of the psychometric function is the PSE, at which reporting of left or right motion was equally likely. The control condition (only translation without rotation) is represented by circles. Conditions in which there was a simultaneous distracting yaw motion are shown with triangles. Data for individual subjects (1–17) are in order of age; error bars represent the 95% CI. Average data from subjects under age 50 (1–12) are shown in the column labeled Y (younger). Average data from the subjects 50 and over (13–17) are shown in the column labeled O (older). For the grouped data, error bars represent SD. A: for-aft direction of surge was reported during yaw rotation. Although some individual subjects had some biases during this condition, the average bias was close to 0. B: the left-right direction of sway was reported during yaw rotation. Most subjects had a bias, such as with leftward yaw, as they were more likely also to report leftward translation.

When the perceived direction of rotation was considered, translation did not reliably bias the results. In the aggregate data, there was no bias in perception of yaw rotation alone or when combined with surge (Fig. 6A) or sway (Fig. 6B) translation. When the effects of the translation distractor were considered across individuals overall, there was no effect of surge (paired t-test, P = 0.7) or sway (P = 0.9) when the results of all subjects were considered. However, within individuals, some large and significant differences occurred. For instance, in the yaw-sway condition, for subject #17 (the oldest in the series) the right-left sway motion had a large influence on how rotation was perceived, but even in individual subjects, when significant biases were found, they were not in a consistent direction. Thus although concurrent yaw rotation that created a significant bias in sway (left-right translation) was perceived (Fig. 5B), the reverse was not the case (Fig. 6B), as a concurrent translation did not bias rotation perception.

Fig. 6.

Yaw rotation bias. The control condition in which only rotation occurred (no translation) is represented by circles. Conditions in which there was a simultaneous distracting translation are shown with triangles. Data for individual subjects (1–17) are in order of age; error bars represent the 95% CI. Average data from subjects under age 50 (1–12) are shown in the column labeled Y (younger). Average data from the subjects 50 and over (13–17) are shown in the column labeled O (older). For the grouped data, error bars represent SD. A: distracting stimulus was surge (for-aft) translation. B: distracting stimulus was sway (lateral) translation. Although some subjects did demonstrate biases, they were not in a consistent direction, and the average bias was near 0.

An additional series of experiments examined the relationship between rotation and translation perception in terms of rotation axis. In this series of experiments, subjects reported their direction of perceived translation, and the axis of rotation was adjusted to find the point at which they were equally likely to report translation in either direction. When the subjects reported their surge (forward vs. backward) direction, and the axis was moved left and right, the PSE occurred when the axis was near the center of the head (Fig. 7A) in most subjects. However, when the subjects reported sway (left vs. right), and the axis was moved forward and backward, the PSE occurred with the axis anterior to the head in all but 2 of the 17 subjects (Fig. 7B). The average axis position across subjects was 28.7 ± 10.5 cm (mean ± SE). Although the anterior displacement of the axis at which the PSE occurred was further anterior in the older group, this was due to the outlier data of the oldest subject (#17).

Fig. 7.

Position of the rotation axis at which subjects were equally likely to report translation in either direction. The rotation direction (left vs. right) is represented using triangles. Data for individual subjects (1–17) are in order of age; error bars represent the 95% CI. Average data from subjects under age 50 (1–12) are shown in the column labeled Y (younger). Average data from the subjects 50 and over (13–17) are shown in the column labeled O (older). For the grouped data, the error bars represent SD. A: the rotation axis was translated laterally, and the subjects reported fore-aft displacement. The axis at which the PSE occurred was located near the midline. B: the rotation axis was translated fore-aft, and the subjects reported the direction of lateral displacement. The axis at which the PSE occurred was anterior to the head (negative) in most subjects but was posterior in 2 young subjects. The eldest subject (#17) had an axis displaced >1 m anterior, which skewed the results for the older group.

The sway axis location, at which translation was no longer perceived (Fig. 7B), was significantly correlated with the bias in sway with a yaw distractor (Fig. 5B; r² = 0.49, P = 0.045). The subject with the furthest anterior rotation axis (#17) also had the largest sway bias. Furthermore, the two subjects where the axis was located behind the head (#3 and #10) also had the bias with sway yaw in the opposite direction of the other subjects.

The results are averaged across conditions and age categories for threshold (Fig. 8) and bias (Fig. 9). The main findings were that combining translation with a yaw distractor had a large effect of increasing thresholds in younger subjects and that a bias occurred when subjects were asked to report sway direction in the presence of yaw rotation (Fig. 9A). Such a bias could be eliminated by moving the axis in front of the head (Fig. 9B).

Fig. 8.

Summary of threshold by age and test condition. Each symbol represents the average data for each age group. Data are grouped by the test stimulus with the distractor in parentheses. Control data are represented as a circle; conditions in which there was a distractor present are represented with a triangle. Error bars represent ± 1 SD.

Fig. 9.

Summary of bias (i.e., PSE) by age and test condition. Each symbol represents the average data for each age group. Data are grouped by the test stimulus with the distractor in parentheses. Control data are represented as circles; conditions in which there was a distractor present are represented with triangles. Error bars represent SD. A: combined translation and rotation. Biases were largest when subjects reported the direction of sway in the presence of a yaw distractor. B: location of rotation axis at which the PSE of translation direction occurs. Yaw rotation to the left and right is shown using triangles.

DISCUSSION

The current study examined both the threshold and bias of combined movements. In situations in which movements are usually not paired during natural activity, such as yaw and surge, the thresholds, similar to these stimuli when they were presented independently (Figs. 3A and 4A), and the biases were near zero (Figs. 5A and 6A). The thresholds of isolated stimuli found here were largely in line with those previously reported. In the subjects under 50, the thresholds for translation (0.8 cm/s) and rotation (0.9°/s) with the 1-Hz stimulus user were similar to those reported in our previous study (Roditi and Crane 2012a) that used a different staircase procedure (2 down, 1 up). Although the older-than-50 subjects had a larger (but not significant) threshold than the younger population in the current study, this finding is in line with the previous report (Roditi and Crane 2012a). Most other studies have not examined age effects and have focused on a young population. However, isolated translation thresholds reported here for younger subjects are also within the range reported by others (Benson and Brown 1989; Benson et al. 1986; Gianna et al. 1996; Gundry 1978; MacNeilage et al. 2010b; Soyka et al. 2011, 2015; Walsh 1961). Similarly, the current rotation thresholds are within the 0.5–1°/s range reported by others (Benson and Brown 1989; Benson et al. 1989; Clark 1967; Grabherr et al. 2008; MacNeilage et al. 2010b; Mallery et al. 2010; Soyka et al. 2015). This agreement supports the validity of the techniques used here. The current work did find an occasional subject with a bias (e.g., a neutral stimulus was reliably identified in 1 direction) in perception of a single modality stimulus (e.g., subject #6 in Fig. 5A), and this behavior has been observed by others (Merfeld 2011), although previous studies have excluded those subjects from the final data set. Because we were interested in bias, these subjects were not excluded from the current study. Furthermore, many prior studies may have provided nonvestibular cues to the absence of a vestibular stimulus due to lack of noise or vibration that might tend to eliminate biases, whereas the current study included some noise and vibration that were delivered during every stimulus, which may have made it more likely to see a bias.

The main innovation of the current work is the measurement of threshold and bias when subjects experienced combined rotation and translation. To our knowledge, this has not been reported previously, and bias has been considered previously in this way. When subjects were asked to draw a path that they experienced, involving both translation and rotation, the experience seemed to be dominated by rotation, which suggested a bias, although it was not explicitly measured (Ivanenko et al. 1997). Other studies that looked for a bias did not find one (MacNeilage et al. 2010b; Soyka et al. 2015). In the Soyka et al. (2015) study, the lack of a bias was not surprising, because rotation and translation always occurred in the same direction, so when there was no translation, there was also no rotation (Soyka et al. 2015). Thus situations that might cause a bias were not tested. In the MacNeilage et al. (2010b) paper, the stimulus used was 2 s in duration, and the rotation and translation stimuli were very different relative magnitudes than those currently reported. In their rotation-detection experiments, the angular velocity was purposely always near the detection threshold at 1°/s, relative to the forward linear velocity, 5–21 cm/s. Similarly, in the translation-detection experiments, the linear velocity was 3 cm/s, and angular velocity was 5–20°/s. The task was also different in that the subjects were only asked to identify if a stimulus was present and did not report direction. Furthermore, the linear stimulus was always in the fore-aft (surge) direction, and even in the current experiments, combined surge and yaw rotation did not lead to biases (Fig. 9A). The current data provide multiple lines of evidence to support concurrent yaw rotation, biasing the perception of sway. First, a yaw rotation consistently shifted the PSE of sway perception, such that lateral translation in the opposite direction had to be added for subjects to perceive the direction of translation as neutral (Fig. 9A). Second, the axis of rotation had to be moved anterior to the head for lateral translation to be perceived as neutral (Fig. 9B).

It is worthy of mention that when yaw rotation and translation occur simultaneously, the peak acceleration of the stimulus and peak deceleration occur in slightly different directions. In these experiments, the total rotation was small at 7°, and the displacement between the peak acceleration and deceleration was smaller at 5.7°. The difference in the magnitude of the acceleration/deceleration vector along the interaural axis was negligible with this displacement [i.e., cosine (5.7°) or 99.5% the same]. Although this could explain a small increase in detection threshold, it is unlikely to explain the much larger effects seen here. During deceleration, the turning of the head also caused the deceleration to have a small, nasal-occipital component (<10% of the lateral stimulus). It seems unlikely that this factor alone biased results, since subjects were asked to report only the lateral component of translation, which was an order of magnitude larger. Furthermore, if effects, due to small changes in the deceleration vector, caused biases, then one would expect to see these in the surge-yaw also, but they were only observed in the sway-yaw condition. Such an effect would not occur when subjects are rotated about a curved path, as occurred in the sway-yaw conditions. Because the effects were consistent between the curved path and sway-yaw conditions, such an effect would also be unlikely to describe the biases seen here.

One possible explanation for concurrent yaw movement biasing the perception of sway is that this might be a mechanism for filtering movements that occur during ambulation so that more atypical movements (e.g., movements that may be a precursor to a fall) are more likely to be perceived. When the axis was moved in the fore-aft direction (Fig. 1), most subjects reported a PSE between left and right translation, with a rotation axis 23 cm anterior to the head (Fig. 7B) in young controls. In older controls, the axis was further anterior (mean 55 cm), but this was largely due to a single elderly subject who had the axis located >1 m out. The anterior axis of rotation at which the PSE occurred was suggestive of the anterior axis previously observed during ambulation (Crane and Demer 1997). This finding suggests that a PSE of left vs. right translation occurs with an anterior axis so that activities outside the typical pattern of motion during ambulation may be more likely to be perceived.

Further quantitative analysis of the previously reported head movements during ambulation (Crane and Demer 1997) can be used to suggest an axis of rotation to compare with the current perceptual estimates. In the horizontal plane [see Fig. 6 in Crane and Demer (1997)], this ratio was ∼1.1 cm/s or 1.1°/s during walking and slightly lower, at 1.0, during running. Both of these estimates were based on the root mean squared velocity. In the vertical plane, the ratio was much lower at ∼0.2 cm/s or 0.2°/s. When rotating about an axis at the center of a circle, the velocity (v) along the circumference is v = 2πrω, where r is the radius to the axis, and ω is the angular velocity in radians/second. The conversion to angular velocity in degrees/second (ϕ) and the solution for the axis of rotation yield is r = 180d/πϕ. Thus the ambulation data suggest an axis of 57 cm for the horizontal plane and 11 cm for the vertical plan. In the current data, the average axis position at which translation was not sensed was 55 cm in the older subjects (approximately the same as the axis estimate using motion in the horizontal plane and 28.7 cm in all of the subjects or approximately the mean of these 2 values). It should also be mentioned that the values previously reported for the ratios of head translation and rotation during ambulation (Crane and Demer 1997) were for an outlying viewing distance, and with more proximal targets, the axis decreased. Our perceptual experiments were done in darkness, and it is unknown if they were viewing distance dependent, but a screen was ∼25 cm in front of the subject. This analysis suggests that the axis of rotation, at which translation perception at the PSE is similar to that used in ambulation, has several limitations, including being measured in a different population of individuals, calculated from only root mean squared velocity.

This paper also measured thresholds of perception during concurrent rotation and translation. Rotation thresholds were unchanged during concurrent translation (Fig. 8), consistent with what was previously reported using a different method (MacNeilage et al. 2010b). The yaw thresholds found here of 0.9°/s (under age 50) and 1.8°/s (over 50) were similar to that found previously for similar frequencies (Benson et al. 1989; MacNeilage et al. 2010b; Miller and Graybiel 1975; Roditi and Crane 2012a; Soyka et al. 2015). Translation perception thresholds were slightly higher with concurrent rotation. As with our prior study, we found high thresholds in subjects over 50 (Roditi and Crane 2012a). Thresholds averaged <1 cm/s in younger subjects and 2–3 cm/s in the older group (Fig. 8). With surge, the thresholds were higher when measured with concurrent rotation, similar to what was previously reported (MacNeilage et al. 2010b). However, the MacNeilage et al. (2010b) study did not examine lateral (sway) translation and rotation, a condition in which we found even more dramatic increases in thresholds, especially in the younger subjects (Fig. 8). The reason for this increased translation threshold when a distracting yaw rotation was present may be due to more noise introduced into the otolith organs by the rotation. Any rotation also includes some translation, since the otolith organs are displaced from the center of the head, and thus when the head rotates, some translation of at least the otolith organs on one side will occur. However, the opposite is not true—translation does not introduce any rotation; thus a distracting translation stimulus does not increase the rotation thresholds. This was true in the current study (Fig. 8), as well as previously published by others (MacNeilage et al. 2010b).

The current study purposely included subjects that encompassed the full age range of the adult human population. One unexpected finding was that six subjects were unable to complete the study, because they perceived only rotation when translation and rotation were combined. These subjects were ages 47–86 (mean 68), and 5 of the 6 were in the over-50 group; these subjects comprised 45% of the individuals older than 50 recruited for this study and 8% of those under 50, suggesting that this was an age-related issue. One previous study suggested that human subjects were more likely to perceive the rotation component with combined rotation and translation (Ivanenko and Grasso 1997), although the ages of the subjects were not reported. In this study, we also found that older subjects generally had higher thresholds, similar to what was previously reported (Roditi and Crane 2012a). Here, we failed to see a statistically significant difference between age groups in most measures. This was likely because there was a lot of intrasubject variation, and we may have selected higher functioning, older individuals in the current study by excluding those that could not report translation during a combined translation-rotation stimulus. It is possible that with a larger study population, significant differences with age could be found; however, the recognition of age differences was not a major goal of the current study, and it would have been of questionable validity, since a large fraction of older individuals seems to be unable to do all of the tasks required. However, the inability to sense translation in the presence of rotation may be another factor that leads to increased fall risk in the elderly (Kannus et al. 1999; Tinetti 2003), in addition to the effect of peripheral vestibular dysfunction (Agrawal et al. 2012; Baloh et al. 2003). Given that the amount of translation is larger in elderly individuals during ambulation (Anson et al. 2013), the increased translation and decreased ability to perceive it when combined with rotation could be major factors in fall risk.

The integration of semicircular canal and otolith signals has also been studied during vestibulo-ocular reflex function. The consensus of these studies is that there is a linear, essentially additive effect of the rotation and translation signals in humans (Crane and Demer 1999; Seidman et al. 2002). However, that is not consistent with the current results in which perception with concurrent rotation and translation is directly predicted by the perception of each alone.

Conclusions.

When rotation and translation occur concurrently, the perception of sway is biased in the direction of rotation, such that during rotation about a head, centered axis subjects are likely to perceive translation in the same direction. This may be a mechanism for filtering perception, such that the translation that occurs during a combination of head movements in ambulation is not perceived. The threshold of sway perception was also higher when rotation occurred simultaneously. In almost one-half of subjects over 50, when rotation was present, the subjects could not perceive any translation. However, distracting concurrent translation did not influence rotation perception.

GRANTS

Research funding for this work was provided by the National Institute on Deafness and Other Communication Disorders (Grants R01 DC013580 and K23 DC011298) and a Triological Society Career Scientist Award.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author.

AUTHOR CONTRIBUTIONS

B.T.C. conception and design of research; B.T.C. performed experiments; B.T.C. analyzed data; B.T.C. interpreted results of experiments; B.T.C. prepared figures; B.T.C. drafted manuscript; B.T.C. edited and revised manuscript; B.T.C. approved final version of manuscript.