Ageing vision

The decline in many visual functions in older adults can be directly attributed to anatomical changes of the eyeball. The anatomical changes reduce the quality of sensory inputs to higher-level visual processing and, as a result, visual functions deteriorate. However, some functional declines cannot be fully explained by anatomical changes. Declines of computational efficiency and compensatory heuristics in the higher-level visual processes are responsible for those functional changes. In this section, we will review anatomical changes of eyeball first and functional declines of vision in general.

Correlation between the decline in visual functions and fall risk

The decline in the visual functions of older adults impairs gait and balance control, and consequently increases fall risk [4, 15, 16]. A large-scale survey study, termed the Blue Mountain Eye Study, examined the relationship between visual functions and fall risk in 3299 older adults [20]. The study found that visual acuity, contrast sensitivity, glare sensitivity, and visual field size were significantly correlated with fall risk. Eye diseases such as a cataracts and glaucoma were associated with falls. The study also confirmed that age, sex, psychotropic drug use, and history of stroke are associated with the number of falls. In this section, we will discuss in detail the effect of a specific visual function on fall risk.

Vection and postural control

Vection is a sensation of illusory self-motion that occurs when an observer is exposed to a visual motion signal. It clearly demonstrates that perception of visual motion signals is directly related to postural balance control. Vection has been reported to occur in various directions: anterior-posterior direction/depth motion [69], left-right direction/circular motion [70], and clockwise-counterclockwise direction/roll motion [71].

The magnitude of vection can be manipulated by several factors. Brandt et al. [70] tested the effect of visual field size on vection using a circular vection stimulus that is produced by placing the subject inside a rotating drum with black and white vertical gratings. When the motion stimulus covers the entire visual field or the peripheral visual field without central vision, subjects reported the sensation of self-motion with a perceived speed that matched the speed of stimulus motion. However, when the motion stimulus only covered the central 30° without peripheral vision, subjects reported no sensation of self-motion. The results suggested that the motion signal in peripheral vision is critical to the perception of circular vection. The effect of the visual field size was also found in roll vection. Allison et al. [71] examined sensation of self-rotating motion (i.e. roll vection) in a tumbling room, in which subject sat in a stationary chair observing a furnished room rotating about the body roll axis. The rotation speed of the tumbling room and the degree of the visual field affected the magnitude of roll vection. When full-field view was allowed, the majority of the subjects reported a complete 360° body rotation. The percentage of subjects experiencing 360° roll vection decreased as the degree of visual field decreased.

Predictably, presenting stationary objects along with vection stimuli can diminish the magnitude of vection. A more important finding was that the effect of stationary objects on vection depends on the relative distance of between stationary objects and moving stimulus. Stationary objects presented closer to subjects than moving stimuli do not affect the magnitude of vection, whereas stationary objects presented further away than moving stimuli markedly reduced the magnitude of vection [72, 125]. The inhibitory effect of stationary objects was consistently found in forward/backwards vection [73]. This result is practical, because closer stationary objects can be perceived to move together with subjects, such as the interior of a moving car.

Vection and postural control are highly correlated as can be demonstrated by the fact that optic flow that generates vection [69] can also induce postural sway of an observer [74, 75]. In addition, the magnitude of vection is correlated with the size of the postural sway in various directions [76–78]. In depth vection direction, the magnitude of vection is positively correlated with postural sway. Postural body sway was measured while optic flow from random dot patterns was presented to participants [76]. Similarly, postural sway was also larger when the magnitude of the perceived roll vection direction was higher [78]. Recently, it was reported that dependence on vision in postural control predicts the strength of vection [79]. In detail, the dependence on vision in controlling upright posture was measured by subjects’ postural sway ratio in opened and closed eyes conditions. The magnitude of perceived vection was assessed by subjective rating given by subjects to vection stimuli that were introduced immediately after the postural control test.

The vection and postural sway correlation can be stimulated by direct visual motion and also by illusory visual motion caused by the motion after effect (MAE) [80]. The postural sway of participants was measured while being stimulated by the MAE. MAE stimulation began with adaptation to left or right motion of a random pixel array. This was followed by a black screen and test pattern which consisted of three conditions: a static random pixel array (expected to produce longer MAE), a dynamic random pixel array in which each pixel was randomly assigned to bright or dark every 16.7 ms (expected to produce shorter MAE), and a blank screen (expected to produce no MAE). MAE was perceived to be significantly longer in the static condition compared to the dynamic condition. Furthermore, postural sway was found to be larger in the static condition than in the dynamic condition.

With age, perceived vection declines but postural sway rate increases. In a study conducted by Haibach et al., postural sway was recorded in two age groups, young adults and older adults, with stimulation in the form of a virtual reality moving room. They were also asked to rate the perceived vection. The results revealed that older adults experienced smaller vection than young adults, but exhibited larger postural sway. The lower rate of perceived vection in older adults suggests that a reduction in proprioceptive feedback due to ageing may contribute to a larger postural sway [81].

Considering that visual motion stimulation is important in vection and postural control, in-depth investigation of visual cue components in this visual motion stimulation will be fruitful in aiding our understanding of fall risk with ageing vision.

Sensory reweighting

In order to effectively estimate and control the state of body posture, the balance control system should integrate sensory inputs from proprioceptive, vestibular, and visual systems [82, 83]. There has been an interesting line of research suggesting that the degree by which the balance control system relies on each sensory input is not hard-wired, but adaptively changes. The phenomenon, which is termed sensory reweighting, suggests that the balance control system will rely on one sensory signal over the other as the sensory environment changes.

Sensory reweighting is observed in various contexts. Ageing alters the relative accuracy of sensory signals, and, consequently, older adults rely more on vision to control postural balance than on vestibular or proprioceptive signals [81]. Loss of vestibular function affects the way patients respond to the visual signals to control posture. Peterka [83] measured the size of postural sway as a function of the magnitude of visual signals. As the amplitude of the visual signal increased, normal subjects’ decreased gain to visual inputs was demonstrative of adaptive sensory reweighting. However, patients with vestibular loss exhibited a constant gain regardless of the magnitude of visual inputs.

Sensory reweighting can be classified as inter- or intra-modality reweighting. Inter-modality reweighting indicates a change in the gain of a sensory modality due to the magnitude change of motion signal of the same sensory modality [82, 84, 85]. Intra-modality reweighting refers to a change in the gain of a sensory modality due to the magnitude change of motion signal from a different sensory modality [84, 85].

An example of intra-modality reweighting was observed in an experiment conducted by Polastri et al. [85]. When an observer was surrounded by visual motion stimulation of constant frequency and standing on a platform that increased the sway amplitude, postural sway decreased relative to the platform. Another example of intra-modality reweighting is demonstrated when a randomly fluctuating visual motion stimulus was presented for 60 s and was introduced before presenting the main visual motion stimulus to observers [82]. The magnitude of sensory sway in response to the main visual motion stimulus was reduced. This result indicates that the postural control system re-adjusts the reliance on sensory modalities in response to the previous experience in an identical context.

Inter-modality reweighting was also demonstrated in the same study conducted by Polastri et al. [85]. When postural sway relative to the platform was decreased by an increase in the platform amplitude, postural sway relative to the visual stimulus increased as a result. Another example can be demonstrated by varied visual motion stimulation and constant touch motion stimulation amplitudes being introduced to observers simultaneously. Increased visual motion stimulation amplitude did not significantly affect the gain from sensory touch. However, when the condition was reversed (i.e. the amplitude of touch motion stimuli was varied and visual motion stimuli amplitude was set to constant), the increased amplitude of touch motion stimuli significantly affected vision sensory gain [84].

Sensory reweighting is slower in older adults when balance is being maintained. Hay et al. [86] investigated the mechanism by which older and young adults maintain their postural balance when the availability of visual and proprioceptive information was manipulated. When the proprioceptive inputs were perturbed by means of tendon vibration, both elderly and young adults exhibited a marked decrease of stability. A clear distinction between older and young adults was found when the perturbed proprioceptive inputs were reinserted. Young adults were able to rapidly integrate available information to control balance, whereas older adults were unable to utilize the reinserted sensory signals. These findings suggest that elderly adults are slower in adjusting weights on sensory modalities in response to their availability. A recent study also supports this result as, unlike young adults, older adults lose the ability to select stimulation such as galvanic vestibular stimulation (GVS) in order to reduce postural instability when perturbation of ankle and vision was introduced [87]. In a recent neuroimaging study on balance control, activity in the brains of older adults was distributed across several locations when somatosensory and visual information was absent, while in young adults the temporal-parietal region was more active, suggesting that in older adults sensory re-weighting demands more attention [88].

It is known that sensory reweighting is important in balance control and that ageing affects this ability even in healthy adults [89]. However, despite of older adults’ responses are generally poorer than young adults’ in a condition where sensory conflicts were presented, 1-h exposure to sensory conflict induced by virtual environment exhibits a better adaptation in older adults’ muscle response. Muscle response delay is shorter in first 10 trials compared to last 10 trials suggesting the possibility of postural balance improvement using sensory reweighting in older adults [90]. These results indicate that knowledge of the mechanism underlying sensory reweighting in older adults may become a stepping stone in understanding falls and in the development of prevention strategies.

Spatial suppression in motion perception

In the “Visual motion perception and balancing” section, we discussed the notion that the effect of ageing on the sensitivity to motion signals depends on the properties of the motion stimulus. The sensitivity to a small, low-contrast motion stimulus deteriorated with age, but the sensitivity to a large, high-contrast motion stimulus improves [42]. In other words, older adults are better at detecting the direction of a large, high-contrast motion stimulus than young adults. It was suggested that the relatively poor performance of young adults in the task was caused by strong centre-surround antagonism in visual motion perception. Centre-surround antagonism in visual motion perception refers to the fact that the firing rate of motion-sensitive neurons in the MT (middle temporal area in the brain) decreases when the surrounding area of the receptive field is stimulated [91]. Thus, the sensitivity to the motion stimulus declines as the stimulus size becomes larger [43, 92].

It is possible that the sensitivity to a small, low-contrast motion stimulus and the sensitivity to a large, high-contrast motion stimulus, which reflect the magnitude of spatial suppression, are related to the fall risks in older adults in a completely different manner. Further investigation of this theory will clarify our understanding of the relationship between visual motion perception and fall risk.

Theories of multisensory integration

Balance control involves several sensory modalities, including visual motion perception. In order to use the multi-modal sensory signals efficiently, the system should be able to integrate visual, vestibular, and somatosensory signals properly. Therefore, understanding the principles by which our balance control system integrates multisensory information is critical to understand human balance control and fall risk. However, as far as we are aware, the ability to integrate multisensory signals has not been discussed in the context of fall risk factors.

There are two well-supported theories of multisensory cue integration that might contribute to the understanding of balance control and fall risk. The first theory states that humans integrate multi-sensory signals in a statistically optimal manner [93]. Statistical optimality of cue integration implies that the perceptual system depends more on sensory signals with low uncertainty than sensory signals with high uncertainty. For example, it is expected that the balance control system relies more on the visual stimuli when the stimulus is clearly presented (e.g. during the daytime) than when the visual stimulus is unclear (e.g. at night). It is largely unknown how the uncertainty of sensory signals affects the sensory integration in balance control.

The second theory is known as the causal inference model of multisensory cue integration [94]. According to the theory, temporal and spatial distances between multisensory signals determine the degree of perceptual integration. When sensory signals from different modalities are spatially and temporally distant, the perceptual system processes the signals independently, and when sensory signals from different modalities are spatially and temporally close, the perceptual system integrates the signals. By applying the two general theories of multisensory integration to understand human balance control, we could better understand why and how the balance control system reacts to diverse patterns of visual, vestibular, and somatosensory signals. Among visual inputs, visual motion signals are highly correlated with the vestibular and somatosensory signals. Thus, multisensory integration is particularly important in understanding how visual motion signals are related to balance control and fall risk.

Older adults who are prone to falls have a tendency to over-integrate multisensory signals [95, 96]. This is demonstrated in the sound-induced flash illusion, in which the number of concurrently presented sounds affects the perceived number of flashes. Subjects perceive two visual flashes when one visual flash and two consecutive auditory stimuli are briefly presented. In general, the delay between visual and auditory stimuli diminishes the effect. However, the older adults who are prone to falls report the illusion even when the delay between visual and auditory stimuli is relatively long. This result suggests that the risk of falls may be associated with the inability to infer the causes of multisensory signals [95].

Understanding the mechanism of multisensory integration in postural control is important in studying fall risk and developing prevention strategies. In a recent study, balance training intervention successfully improved postural balance control in healthy and fall-prone older adults, and the improvement was demonstrated to correlate with multisensory processing efficiency [97].

Cognitive load and balance control

Many studies have examined the effect of cognitive load on postural control [98–103]. Cognitive spatial processing and postural balance control may rely on the same neural mechanics. When participants performed spatial and non-spatial memory tasks in sitting and standing positions in the Romberg test, a significantly higher number of errors were observed in the spatial memory task in the standing position compared to the sitting position. However, in the non-spatial memory task, the number of errors was not significantly different in the standing and sitting positions [104]. Postural sway was also affected by a secondary cognitive task [105]. Young and older adults without a history of falls exhibited significantly larger postural sway in a sentence completion task. However, older adults with a history of falls exhibited significantly larger postural sway in both the sentence completion and perceptual matching tasks. Moreover, the study also found that the surface condition (fixed vs swaying) had no significant effect in either age groups. The results suggested that the task of postural control shared cognitive resources with other cognitive tasks and that fall risk in older adults was related to a lack of cognitive resources. Similar results were also observed when cognitive task was demanding attention [106]. Participants were instructed to verbally classify an auditory tone as being high or low during several conditions of balance tasks: a combination of two ground conditions, sway and fixed, and three visual conditions, open, closed, and visual motion. In healthy older adults, postural balance control was affected when visual and somatosensory information was removed simultaneously. However, in balance-impaired older adults, the auditory task affected postural balance control in all sensory conditions. In contrast, Swan et al. [103] demonstrated that adding cognitive load decreased postural sway. In their study, subjects were required to stand still while performing cognitive tasks under different visual (open vs closed eye) and ground (fixed vs swaying pedal) conditions. Cognition tasks were spatial or non-spatial memory tasks. Surprisingly, the results indicated that in the closed-eyes/swayed-ground condition, which was the most difficult condition, cognitive tasks reduced the postural sway. This effect only appeared in the older adults group. Using the same memory task as one of the secondary tasks, Bergamin et al. [107] also observed improvement in postural stability in older adults by adding visuospatial tasks while maintaining balance. Recently, a cognitive task was also reported to decrease postural sway in young adults [108, 109]. Diverting attention from postural control decreases postural sway during the continuous cognitive task but not during the irregular cognitive task [108]. The continuous cognitive task in this study consisted of mentally counting and summing a series of three-digit numbers and mentally performing a series of simple mathematical equations. The irregular cognitive task consisted of verbally responding to random auditory stimuli and distinguishing high-pitched beeps from low-pitched beeps by verbally responding only to the high-pitched beeps [108]. In line with the findings of the studies mentioned above [103, 107–109], it has been shown that postural sway is smaller when subjects focus attention on a cognitive task compared to when subjects focus attention on postural control itself. In the study, participants were asked to silently count numbers in the cognitive task condition, and participants were asked to minimize movement of the hips in the postural control condition [110].

Although inconsistent findings regarding the effect of cognitive load on balance control necessitate further investigation, it seems clear that maintaining posture requires significant cognitive resources. We suggest that one of the main reasons for the sizable effect of cognitive load on balance control is that visual motion processing requires cognitive resources. The motion signal detected from a retinal image is an ambiguous signal. Retinal motion can represent movements of the head, eye, or objects in the world, and identifying the causes of retinal motion signals in sufficiently short duration would require sizable cognitive resources. Indeed, it has been demonstrated that the ability to detect briefly presented visual motion signals is a good indicator of IQ [111]. We suggest that further research on the relationship between cognitive load, visual motion perception, and fall risks may reveal one of the key factors responsible for the falls in older adults.