Comparing the impact of the method of adjustment and forced-choice methodologies on subjective visual vertical bias and variability

Abstract

BACKGROUND:

Previous research suggested that the method of adjustment and forced choice variants of the subjective visual vertical (SVV) produce comparable estimates of both bias and variability. However, variants of the SVV that utilize a method of adjustment procedure are known to be heavily influenced by task parameters, including the stimulus rotation speed, which was not accounted for in previous SVV research comparing the method of adjustment to forced-choice.

OBJECTIVE:

The aim of the present study was to determine if (1) the SVV with a forced-choice procedure produces both bias and variability estimates that are comparable to those obtained using a method of adjustment procedure, (2) to see if rotation speed impacts the comparability of estimates and (3) quantify correlations between the estimates produced by different procedures.

METHODS:

Participants completed a variant of the SVV which utilized a forced-choice procedure as well as two variants of the SVV using a method of adjustment procedure with two different rotation speeds (6°/s and 12°/s).

RESULTS:

We found that the bias estimates were similar across all three conditions tested and that the variability estimates were greater in the SVV variants that utilized a method of adjustment procedure. This difference was more pronounced when the rotation speed was slower (6°/s).

CONCLUSIONS:

The results of this study suggest that forced-choice and method of adjustment methodologies yield similar bias estimates and different variability estimates. Given these results, we recommend utilizing forced-choice procedures unless (a) forced-choice is not feasible or (b) response variability is unimportant. We also recommend that clinicians consider the SVV methods when interpreting a patient’s test results, especially for variability metrics.

Keywords

Subjective Visual Vertical Gravitational Vertical Vestibular Visual-Vestibular Integration

1 Introduction

While knowing where one’s body is relative to gravity may seem trivial, the inability to do so can potentially result in injury [8]. There are a plethora of different conditions that may impact one’s ability to determine where their body is relative to gravity, including an array of different vestibular disorders, making tests that accurately assess this ability desirable [10]. One of the most widely utilized tests of this ability, in both clinical and research settings, is the Subjective Visual Vertical (SVV) task [1, 24].

The SVV is a perceptual task that assesses visuo-vestibulo-somatosensory integration by having participants judge the perceived tilt of a visual line relative to gravity [3 , 16]. There are a variety of different procedures which can be employed for the SVV test. However, the most commonly utilized procedure is a psychophysical procedure called the Method of Adjustment [21]. This procedure entails the participant (i.e. the observer) moving a stimulus to a position that matches their internal representation of verticality either manually (via a keyboard or other human interface) or via verbal instructions to the experimenter [11].This allows one to estimate the noise magnitude (quantified as the standard deviation) as well as bias (i.e. the quantity and direction of the deviation of their judgements from the true vertical) [1]. While this procedure has been widely utilized, it is not without limitations. Specifically it has been found that results can be impacted by extraneous factors including the starting position of the stimulus, the stimulus length, the stimulus type, how one indicates their response and rotation speeds (i.e. the speed at which the adjustments are made by the observer) [1 , 25].

In an attempt to overcome the aforementioned issues, several studies have utilized a forced-choice procedure with two choices [1 , 21]. This procedure entails presenting a stimulus and asking the participant to determine if is tilted clockwise or counter clockwise from vertical [2]. As the angle offset from vertical decreases, the participants’ performance becomes more variable and less accurate. Using this method one can compute a full psychometric function meaning one can compute a threshold (i.e. the range of values for which the participants’ verticality judgements are uncertain and is indicative of noise magnitude) and the bias (i.e. the amount and direction of the deviation of their judgements from the true vertical) [1].

There is a dearth of empirical data assessing the comparability of the forced-choice procedure and the method of adjustment procedure. However, a recent study conducted by Baccini and colleagues [1] suggests that forced-choice and method of adjustment procedures yield qualitatively similar bias and variability estimates. Furthermore, it was found that performance on the measures correlated with each other. However, as noted above there are a variety of factors that can impact performance on the SVV when using the method of adjustment including the rotation speed of the stimulus, which was not manipulated or precisely controlled in Baccini and colleagues [1]. This is problematic because past research suggests rotation (even in the absence of visual orientation cues) can impact verticality perception [9]. There is some evidence that suggests this effect is, at least in part, the result of ocular torsion which is more pronounced when the movement of the rotating stimulus is controlled by the participant [12, 20]. Thus, it remains unclear if the comparability of these procedures persists when one precisely controls and varies the rotation speed of the stimuli or if the same rotation speed is acceptable for all individuals (e.g., (a) both old and young adults, or (b) healthy subjects versus patients with a variety of diagnoses).

In the present study we directly compared the SVV with the forced-choice procedure to two SVVs which utilized the method of adjustment procedure; one with rotation speeds of 6°/s and the other with a rotation speed of 12°/s. It was hypothesized that the variability and bias estimates produced by the different procedures would not differ (as is suggested by previous studies Baccini et al. [1]). Positive correlations between the variability estimates and bias estimates were also expected (based upon the findings of Baccini et al. [1]).

2 Method

2.1 Participants

A total of 10 healthy participants (4 female, 6 male) were recruited from a large state university. The average age of the participants was 26.4 years and ranged 20–35 years. All participants had normal or corrected to normal vision. Each participant gave written consent in accordance with the Declaration of Helsinki. The study protocol was approved by the University of Arizona Institutional Review Board.

In order to determine if there were any outliers Z Scores were computed for each participant’s bias and standard deviation estimates for all SVV variants. Having a Z score greater than 3 or less than –3 for any of the SVV variants was grounds for exclusion. No participants met the exclusion criteria and thus all were included in the analyses.

2.2 Stimulus presentation apparatus

The visual stimuli were displayed on a 24-inch computer monitor (Asus VG248QE), which was viewed through a round window with a 20° viewing angle at a distance of 85 cm from the eyes. The stimulus presentation apparatus was set up in a light proof room, and the monitor’s refresh rate was 60 Hz.

2.3 Visual stimuli

The visual stimulus utilized in both SVV procedures was a Gabor patch which was generated in Matlab (version 2017a) using Psychtoolbox [4]. The Gabor patch had a visual angle of 7° diameter [18] with 2 cycle/° and 80% contrast [1]. This stimuli was utilized to minimize pixelation that may occur during stimuli adjustment. Each Gabor patch was followed by a visual masker (i.e. a bullseye) without orientation cues to minimize potential after effects of the stimuli as was done in Lim et al. [16, 17].

2.4 Method of adjustment subjective visual vertical

When performing the SVV with the method of adjustment procedure participants were asked to rotate a Gabor patch, in the center of the computer screen, to what they perceived to be vertical. This was accomplished by the observer pressing the left and right arrow keys on a keyboard (with their right hand) to rotate the Gabor patch. Observers often continuously pressed a button until the patch approached appearing vertical and then switched to a discrete button press strategy. Once satisfied, they indicated their final answer by pressing the up arrow key. Upon doing this a visual masker without orientation cues (i.e. a bulls eye) was displayed indicating their answer had been submitted. Then, after 2,000 ms, a fixation point appeared indicating the start of the next trial. The starting position of the Gabor patch alternated between 20 degrees clockwise and 20 degrees counter clockwise.

Participants completed two variants of the method of adjustment SVV procedure. One variant featured a rotation speed of 6°s per second and the other featured a rotation speed of 12°s per second, which occurred each time they pressed the right/left key. The speed was set by the experimenter prior to the beginning of each test. Participants completed a total of 50 trials for each of the two variants. Each variant took approximately 7 to 9 minutes to complete. See Fig. 1 for a depiction of the course of a trial using a method of adjustment procedure.

2.5 Forced-choice subjective visual vertical

When performing the SVV with the forced-choice procedure participants were asked to make judgements about whether the Gabor patch - displayed in the center of the screen for 800 ms. - was rotated clockwise or counter clockwise relative to the gravitational vertical. Once they indicated their answer by pressing the left or right arrow key on a keyboard with their right hand, a visual masker without orientation cues (i.e. a bulls eye) was displayed indicating their answer had been submitted. Then, after 2,000 ms, a fixation point appeared indicating the start of the next trial. The rotation angle of the stimulus was selected by the computer using an adaptive 4 down 1 up staircase, meaning participants had to answer correctly for four consecutive trials for the algorithm to select smaller angles (half of the previous angle) while an incorrect answers yielded selection of a larger angle (i.e. the absolute value of the previous angle minus the starting angle magnitude of 16 degrees]). Participants completed a total of 100 trials over the course of approximately 4 to 6 minutes. See Fig. 1 for a depiction of the course of a trial using our forced-choice procedure and Fig. 2a for an example of stimulus values selected by a 4 down 1 up staircase as well as the responses for each stimulus.

Fig. 1

The course of events in each trial of the SVV Variants. Both method of adjustment procedures began each trial with a fixation point that was on the screen for 2,000 ms followed by the appearance of a Gabor Patch which was visible until the answer was submitted. This was followed by a visual masker (i.e. a bulls eye) that was visible for 2,000 ms until the next trial began. The forced-choice procedures began each trial with a fixation point that was on the screen for 2,000 ms followed by the appearance of a Gabor Patch which was visible for 800 ms. After submitting their answer, an annular visual masker (i.e. a bulls eye without orientation cues) was visible for 2,000 ms until the next trial began.

Fig. 2

(A) An example of the stimulus values and responses for each trial where the stimulus angle was selected using a 4 down 1 up staircase. (B) An example of a plotted psychometric function. The bias value (-0.08) is depicted as a downward facing, blue arrow and represents the point at which the individual was equally likely to judge the line as being tilted to the right or to the left. The threshold value (i.e., 0.95; denoted as σ) is the standard deviation of the participant’s judgements. The blue dots represent the proportion of the responses where the participant indicated that the Gabor patch was tilted towards the right. Note the data in both plots came from the same participant.

2.6 Procedure

Upon providing informed consent, participants were instructed to rest their chin on a chin bar and look at a computer screen, that was 85 cm from their eyes, through a round window with a 20° viewing angle. Participants were then asked to complete the SVV using the; (1) forced-choice procedure, (2) the method of adjustment procedure with a rotation speed of 6° per second and (3) the method of adjustment procedure with a rotation speed of 12° per second. Each test took approximately 4 to 9 minutes (see above for SVV variant specific completion times). The order of the SVV tasks were pseudo-randomly shuffled to eliminate potential order effects. Upon completion of all SVV variants the participants were thanked for participating in the study and were debriefed.

2.7 Data analysis

Data collected from the participants was utilized to estimate the task standard deviation and bias for each of the SVV procedures. For SVV variants that utilized the method of adjustment procedure, these were calculated as the standard deviation and average of each subjects response (i.e. the bias). For the SVV variant that utilized the forced-choice procedure, data were fit using a psychometric function. This was done by fitting these data with a sigmoidal function using maximum likelihood methods [6]. The midpoint between the upper and lower bounds of a psychometric function, i.e. the point at which participants were equally likely to judge the Gabor patch as rotated clockwise and counter clockwise, was used to quantify bias. After calculating the aforementioned parameters in Matlab (version 2017a), the data were entered into SPSS (version 28.0. Armonk, NY: IBM Corp).

3 Results

3.1 Impact of the SVV procedure on variability estimates

To evaluate whether the SVV variant impacted variability (i.e. standard deviation) estimates, a repeated measure ANOVA was performed. It was found that the standard deviation produced by the SVV variants did indeed differ, F(2, 18) = 12.070, p < 0.001, $η_{p}^{2}$ = 0.573. Post hoc LSD tests were utilized to better characterize the differences between the conditions. It was found that the average standard deviation for the method of adjustment 6°/s SVV (M = 2.40°, SD = 0.84°) was higher than the standard deviation found for both the method of adjustment 12°/s (M = 1.83°, SD = 0.55°) and forced-choice (M = 1.11°, SD = 0.52°; p < 0.05) SVV variants. Furthermore, the 12°/s standard deviation was higher than the standard deviation (representing the standard deviation of the noise) for the forced-choice SVV (p < 0.05; see Fig. 3)

Fig. 3

A Box plot depicting the difference between the estimated standard deviation for each SVV variant. Note that the individual data points shown for the forced choice task did not meet the criteria to be considered an outlier (i.e. a Z score of +/–3).

3.2 Impact of the SVV procedure on bias

To evaluate whether the SVV variant impacted bias (μ), a repeated measures ANOVA was performed. It was found that there was not a main effect of SVV variant, F(2, 18) = 1.422, p = 0.267, $η_{p}^{2}$ = 0.136. In other words the bias did not significantly differ between the forced-choice (M = –0.37°, SD = 0.49°), the method of adjustment 6°/s (M = –0.76°, SD = 0.90°) or the method of adjustment 12°/s (M = –0.76°, SD = 0.90°) SVV variants (see Fig. 4).

Fig. 4

A box plot representing the bias values for each SVV variant.

3.3 Correlation between SVV measures

In order to determine if there was a correlation between the different SVV variants’ standard deviation and bias estimates, Spearman correlations were computed. It was found that the bias estimates were correlated with each other (though the correlations between the forced-choice SVV variant and each of the two method-of-adjustment variants fell just short of statistical significance; p = 0.074 and .067). Furthermore, it was found that there was a positive correlation between the variability estimates produced by the method of adjustment procedures that almost reached the level of statistical significance (p = 0.060). However, there was no association between the method of adjustment and forced choice variant estimates (p = 0.651 and .960). Table 1 shows all correlation data (ρ and p values) and Fig. 5 shows a graphic depiction of these data.

Fig. 5

Scatter plots representing the associations between SVV variants’ bias (A-C) and variability estimates (D-F). Note that the line in each scatter plot represents a linear regression fit line.

Table 1

The ρ and p values for each of the spearman correlations

	Forced-choice Bias	Method of Adjustment 6°/s Bias	Method of Adjustment 12°/s Bias	Forced-choice Threshold (Standard Deviation)	Method of Adjustment 6°/s Standard Deviation	Method of Adjustment 12°/s Standard Deviation
Forced-choice Bias	ρ= 1.00
Method of Adjustment 6°/s Bias	ρ= .588p = .074	ρ= 1.00
Method of Adjustment 12°/s Bias	ρ= .600p = .067	ρ= .673p = .003	ρ= 1.00
Forced-choice Threshold (Standard Deviation)	ρ= .006p = .987	ρ= – 0.297p = .405	ρ= – 0.394p = .260	ρ= 1.00
Method of Adjustment 6°/s Standard Deviation	ρ= .345p = .328	ρ= .212p = .556	ρ= – 0.164p = .651	ρ= – 0.164p = .651	ρ= 1.00
Method of Adjustment 12°/s Standard Deviation	ρ= .394p = .260	ρ= .479p = .162	ρ= .261p = .467	ρ= .018p = .960	ρ= .612p = .060	ρ= 1.00

4 Discussion

The aims of the present study were to (1) establish if forced-choice and method of adjustment SVV procedure produces comparable bias and variability estimates, (2) establish whether rotation speed of the stimuli when using the method of adjustment procedure for the SVV impacted the bias and variability estimates and (3) to establish if the different estimates correlated with one another. To this end, each participant separately completed three SVV tasks, one which utilized a forced-choice procedure and two which utilized a method of adjustment procedure. Based upon the results of this study, it would seem that the bias estimates produced by these procedures are comparable at the group level, but the variability estimates are not. This suggests that clinicians and researchers should carefully consider which parameters they are interested in prior to performing assessments. If variability is of interest, utilizing a forced choice variant of the SVV appears likely to be beneficial. Furthermore, this also suggests that clinicians need to account for the SVV variant when interpreting a patient’s test results, especially any utilizing a variability metric.

4.1 Bias Estimates

For our forced-choice task, bias is a term which refers to the point at which participants respond at chance level (i.e. they select clockwise and counter clockwise 50% of the time) –sometimes referred to as the point of “subjective equality”. For our method of adjustment task, bias refers to the mean effect. As was expected the bias estimates produced by all three variants of the SVV did not differ. This finding is consistent with previous research conducted by Baccini and colleagues [1]. Thus it would seem that if one is interested in studying SVV bias exclusively, forced-choice and method of adjustment procedures may both be viable options.

4.2 Variability (i.e., “Threshold”) estimates

For our forced-choice task, threshold is a term which (depending on the procedure utilized) often refers to the standard deviation of the psychometric function/responses (sometimes referred to as psychometric “spread”) [5, 19]. For our method of adjustment task, standard deviation follows the standard statistical calculation of a standard deviation found following repeated trials. Contrary to what was expected, a difference between the procedures was found. Namely, it was found that the standard deviation estimates were larger for the method of adjustment procedure than the forced-choice procedure. Furthermore, when comparing the standard deviation produced by the SVVs which utilized the method of adjustment procedure, it was found that the slower rotation speed (i.e. 6°/s) produced higher standard deviation estimates than faster rotation speeds (i.e. 12°/s). In other words, faster rotation speeds seemed to result in less variability.

The difference in standard deviation estimates is not entirely consistent with past research (i.e. Baccini et al. [1]). However, this is likely explained by methodological differences between the present study and Baccini et al. [1]. One methodological difference that may explain this discrepancy is who adjusted the stimuli. In Baccini et al. [1] the experimenter was the one to adjust the stimuli based on the participant’s verbal responses. Whereas in the present study the participants adjusted the stimuli themselves. Past research suggests that moving lines can elicit ocular torsion [20] that can impact verticality perception [12] particularly when the movement is controlled by the participant. Thus, this may explain why both method of adjustment tasks resulted in more variable verticality judgments than the forced choice SVV variant (which did not rotate). The difference between the method of adjustment SVV variants, in the present study could potentially be the result of the fact that it took more button presses and/or more time when performing the 6°/s SVV variant than the 12°/s SVV variant to get the Gabor patch to the (approximate) true vertical. In other words additional button presses and/or more time could result in additional variability.

Another methodological difference between Baccini et al. [1] and the present study is that the former study only had participants complete 10 trials while using the method of adjustment procedure, whereas in the present study participants completed 50 (per rotation speed). Clearly, more research is needed before strong conclusions can be drawn about the number of trials needed to get good variability estimates. Regardless of the exact reason for the difference reported, it would seem that using a forced-choice procedure for the SVV may be best when the study would benefit from both standard deviation and bias estimates.

4.3 Correlations between estimates

Based upon the findings of Baccini et al. [1], it was expected that the bias and variability estimates produced by each of the three SVV variants would correlate with each other. Contrary to what was expected, only the bias estimates were found to correlate with each other. Though it should be noted that the correlations between the bias values for the forced-choice SVV variant and each of the two method-of-adjustment variants appeared interesting (p = 0.074 and p = 0.067), but did not reach the level of statistical significance for the small population reported herein. Additionally, it was also found that there was a positive correlation between the variability estimates produced by the method of adjustment procedures that almost reached the level of statistical significance. However, there was no association between the method of adjustment and forced choice variability estimates.

While the pattern of correlations described above is contrary to previous research (i.e. Baccini et al. [1]) this finding is consistent with the pattern of results described in sections 4.1 and 4.2. Thus, we suspect that this pattern of findings may be the result of methodological differences between the present study and Baccini et al. [1]. As noted above, in Baccini et al. [1], the experimenter adjusted the stimuli based on the participant’s verbal responses –whereas participants adjusted the stimuli themselves in the present study. This methodological difference is noteworthy because past research suggests that moving lines can impact verticality perception [12] particularly when the movement is controlled by the participant. As noted above, not only were the participants in the present study controlling the rotation of the stimuli in the method of adjustment SVV, but the forced choice SVV also did not feature stimuli rotation. Thus, the larger more variable, variability estimates produced by the method of adjustment SVV variants may be explained by impacts of visual motion on human perception.

4.4 Limitations and Future Directions

It should be noted that the present study was not a large sample (N = 10). This could have resulted in insufficient power to detect a relationship between the estimates produced by the method of adjustment and forced-choice procedures. But the clear significant differences reported for variability suggest that this is not a major concern. Additionally, participants only completed the SVV tasks while sitting upright meaning that it remains unclear whether these results generalize to other body and/or head orientations which can impact performance on the SVV [13, 23]. Furthermore, the method of adjustment SVV variant only utilized two starting angles (i.e. 20° and –20°), meaning it is not known whether the same effect would be found with different starting angles. It should also be noted that there was a limited participant age range and health variability, meaning it is not clear if these results generalize to other age groups and individuals with different health conditions.

Additional research will allow for a better understanding of the estimates produced by forced-choice and method of adjustment methodologies. One open question is whether the findings reported in this study are consistent when different feedback methods are utilized (e.g. keyboard vs joy stick). It will also be important to determine if rotation speed impacts variability estimates in different age groups (e.g. adolescents and older adults). Determining if these findings persist when participants have different body/head orientations and in the presence of physiological perturbations (e.g. neck muscle vibration [14]) will also be important. Additionally, it will be important to collect normative data for the SVV utilizing a forced-choice procedure. As noted above, the forced-choice SVV procedure has not been as widely utilized as the method of adjustment procedure [1 , 21].

5 Conclusion

In sum, the results of the present study suggest that SVVs utilizing forced-choice and method of adjustment procedures, produce comparable estimates of bias but not variability (i.e., standard deviation) at the group level. A consistent pattern of results was also evident when examining correlations, though several did not reach the level of statistical significance. Furthermore, it seems that variability estimates produced by the method of adjustment procedure are impacted by the rotation speed of the visual stimuli. Namely, it seems that slower rotation speeds (i.e. 6°/s) produce more variable estimates than faster rotation speeds (i.e. 12°/s). More research is needed with larger sample sizes and more age groups before firm conclusions about the extent of the generalizability of these results can be drawn.

References

Baccini

, Paci

, Del Colletto

, Ravenni

, Baldassi

The assessment of subjective visual vertical: Comparison of two psychophysical paradigms and age-related performance, Attention,Perception,& Psychophysics 76(1) (2014), 112–122.

Barnett-Cowan

, Dyde

, Fox

, Moro

, Hutchison

, Harris

Multisensory determinants of orientation perception in Parkinson’s disease, Neuroscience 167(4) (2010), 1138–1150.

Barra

, Marquer

, Joassin

, Reymond

, Metge

, Chauvineau

, Pérennou

Humans use internal models to construct and update a sense of verticality, Brain 133(12) (2010), 3552–3563.

Brainard

, Vision

The psychophysics toolbox, Spatial Vision 10(4) (1997), 433–436.

Bremova

, Caushaj

, Ertl

, Strobl

, Böttcher

, Strupp

, MacNeilage

Comparison of linear motion perception thresholds in vestibular migraine and Menière’s disease, European Archives of Oto-Rhino-Laryngology 273(10) (2016), 2931–2939.

Chaudhuri

, Merfeld

Signal detection theory and vestibular perception: III. Estimating unbiased fit parameters for psychometric functions, Experimental Brain Research 225(1) (2013), 133–146.

Clemens

, De Vrijer

Selen

, Van Gisbergen

, Medendorp

, Multisensory processing in spatial orientation: An inverse probabilistic approach, Journal of Neuroscience 31(14) (2011), 5365–5377.

Dakin

, Rosenberg

Gravity estimation and verticality perception, Handbook of Clinical Neurology 159 (2018), 43–59.

Dichgans

, Held

, Young

, Brandt

Moving visual scenes influence the apparent direction of gravity, Science 178 (1972), 1217–1219.

10.

Dieterich

, Brandt

Perception of verticality and vestibular disorders of balance and falls, Frontiers in Neurology 10 (2019), 172.

11.

Fraser

, Makooie

, Harris

The subjective visual vertical and the subjective haptic vertical access different gravity estimates, PLoS One 10(12) (2015), e0145528.

12.

Goonetilleke

, Mezey

, Burgess

, Curthoys

On the relation between ocular torsion and visual perception of line orientation, Visual Research 48 (2008), 1488–1496.

13.

Jaggi-Schwarz

, Ortega

, Hess

Reciprocal error behavior in estimated body position and subjective visual vertical, Experimental Brain Research 150(1) (2003), 122–125.

14.

Kawase

, Maki

, Takata

, Miyazaki

, Kobayashi

Effects of neck muscle vibration on subjective visual vertical: Comparative analysis with effects on nystagmus, European Archives of Oto-Rhino-Laryngology 268(6) (2011), 823–827.

15.

Kheradmand

, Winnick

Perception of upright: Multisensory convergence and the role of temporo-parietal cortex, Frontiers in Neurology 8 (2017), 552.

16.

Lim

, Wang

, Merfeld

Unbounded evidence accumulation characterizes subjective visual vertical forced-choice perceptual choice and confidence, Journal of Neurophysiology 118(5) (2017), 2636–2653.

17.

Lim

, Wang

, Merfeld

Frontal scalp potentials foretell perceptual choice confidence, Journal of Neurophysiology 123(4) (2020), 1566–1577.

18.

Lopez

, Mercier

, Halje

, Blanke

Spatiotemporal dynamics of visual vertical judgments: Early and late brain mechanisms as revealed by high-density electrical neuroimaging, Neuroscience 181 (2011), 134–149.

19.

Merfeld

Signal detection theory and vestibular thresholds: I. Basic theory and practical considerations., Experimental Brain Research 210(3) (2011), 389–405.

20.

Mezey

, Curthoys

, Burgess

, Goonetilleke

, MacDougall

Changes in ocular torsion position produced by a single visual line rotating around the line of sight—Visual “entrainment” of ocular torsion, Visual Research 44 (2004), 397–406.

21.

Paci

, Matulli

, Megna

, Baccini

, Baldassi

The subjective visual vertical in patients with pusher behaviour: A pilot study with a psychophysical approach, Neuropsychological Rehabilitation 21(4) (2011), 539–551.

22.

Pagarkar

, Bamiou

, Ridout

, Luxon

Subjective visual vertical and horizontal: Effect of the preset angle, Archives of Otolaryngology–Head & Neck Surgery 134(4) (2008), 394–401.

23.

Tamura

, Wada

, Kurita

, Matsunobu

, Inui

, Shiotani

Visual effects on the subjective visual vertical and subjective postural head vertical during static roll-tilt, Laryngoscope Investigative otolaryngology 2(3) (2017), 125–130.

24.

Wada

, Yamanaka

, Kitahara

, Kurata

Effect of head roll-tilt on the subjective visual vertical in healthy participants: Towards better clinical measurement of gravity perception, Laryngoscope Investigative Otolaryngology 5(5) (2020), 941–949.

25.

Zakaria

, Salim

, Tahir

, Zainun

, Mohd Sakeri

, The influences of age, gender and geometric pattern of visual image on the verticality perception: A subjective visual vertical (SVV) study among Malaysian adults, Clinical Otolaryngology 44(2) (2019), 166–171.