Increased demand for advanced tactile equipment along with effective haptic languages to convey information calls for basic psychophysical investigations of mechanisms underlying the sense of touch. Assessing spatial acuity will contribute to more efficient tactile applications, for example by determining how closely tactors can be placed for better information transmission. We assessed spatial acuity for vibrotactile stimulation in the lower thoracic region for three different tactor types at two inter-tactor distances, for vertical and horizontal presentation, and compared accuracy in the spine area with the peripheral area. Our main incentives were to gain increased understanding of vibrotactile sensitivity for different tactor types, with the aim of raising awareness of potential differences in outcome when different tactor types are used for vibrotactile spatial acuity studies, and, but also to formulate guidelines for the design of tactile displays.
Our results indicate that vibrotactile discrimination accuracy differs substantially by tactor type with higher accuracy for the N ERMs than for the other two tactor types, and higher accuracy for the P ERMs than the LRAs. The findings are mostly consistent with the specific characteristics of each tactor type. Due to their cylindrical shape, the contact area of the N ERMs depends on how firmly they are pressed against the skin, varying between 125 and 250 mm2. Their contact area is 2.5–5 times larger than those of the other two tactors (50.24 mm2), and larger contact areas have been found to produce higher sensitivity (Morioka et al. 2008). The force, as related to the perceived intensity of a tactor, depends on the interaction between mass, frequency and acceleration, whereby frequency and acceleration reinforce one another (Bolanowski et al. 1994; Morley and Rowe 1990). In line with the results, the mass of the N ERMs (4.6 G) was about four times higher than that of the P ERMs (0.8 G) and LRAs (0.95 G), and the acceleration of the N ERMs (4.0 G) was highest, four times higher than that of the P ERMs (1.0 G) and about three times higher than that of the LRAs (1.4 G). The N ERMs run at a lower load frequency than the P ERMs (62 Hz vs. 132 Hz). Note, however, that the effects of frequency variation on localization accuracy are typically small (Cholewiak et al. 2001; Cholewiak and McGrath 2006). Whether the frequency–acceleration relation (Bolanowski et al. 1994; Morley and Rowe 1990) increases spatial acuity is not entirely clear. For instance, higher acceleration results in stronger surface waves, travelling further from the origin (Franke 1951), which should decrease spatial acuity. Surprisingly, accuracy was lowest for the LRAs, although they can be controlled most precisely, have a similar contact area and mass as the P ERMs, a slightly higher acceleration than the P ERMs, and their load frequency is 2–4 times higher (263 Hz) than of the other tactors. The latter is in line with findings that frequency does not have a strong effect on vibrotactile spatial acuity (Cholewiak et al. 2001). Instead, the difference in accuracy might reflect the way the vibration is generated. The N ERMs generate a complex vibration pattern with rotation on both ends of the tactor causing both perpendicular motion (toward and away from the skin’s plane) and motion parallel to the skin’s plane. Such multiple vibration stimulation might increase perceived intensity and thereby facilitate discrimination perception. LRAs, on the other hand, create force by a magnetic mass attached to a spring and driven by a voice coil (Precision 2018d), resulting in vibration exclusively directed perpendicular to the skin’s surface, which may confine the vibrations and cause smaller surface waves. While lesser vibration spread should improve localization, it might also lower perceived intensity. Azadi and Jones (2014) found that, if put under load, LRAs tend to show a stronger decrease of mechanical input delivered to the skin than other tactor types, possibly affecting the user’s ability to detect the tactile input. In fact, participants in our study reported that it was difficult to discern differences between the LRAs. Future studies should focus on investigating the nature of vibration created by LRAs compared to eccentric mass-based tactors.
When relating the current results to previous studies, our findings for the P ERMs complement the results of Jóhannesson et al. (2017), who found that P ERMs (10 mm diameter), in the same rPL task could be placed as close as physically possible (13 mm c/c) leading to 64% discrimination accuracy. With smaller P ERMs in the current study (8 mm diameter), the inter-tactor distance could be decreased to 10 mm c/c, but even though participants were still able to discriminate two adjacent tactors, accuracy dropped to 45%. Overall, the relatively low accuracy found for P ERMs and LRAs seems to accord well with their small size (as related to force), and the very small tactor distance of only 2–3 mm in between them (10 mm center-to-center), indicating approximation of the threshold of vibrotactile discrimination acuity. Notably, however, the accuracy for the N ERMs was higher than in the results of Novich and Eagleman (2015), who tested the same N ERMs in the same body area using the 2PT method with either spatial stimuli (single motor) or spatiotemporal stimuli (sweeps of two motors). They reported that accuracy was only higher than chance at a tactor distance of 40 mm. In our study, we constrained the N ERMs to 62 Hz load frequency because preliminary tests revealed that participants felt uncomfortable when we ran them at 120 Hz, or higher. In our 3AFC task using the rPL method and spatial stimuli, the accuracy for N ERMs was higher than chance at a 20 mm c/c distance, with accuracy rates of 65% (53%) for horizontal (vertical) presentation. A possible explanation for the lower accuracy in Novich and Eagleman (2015) is that the tactors were run at a high frequency (340 Hz) and acceleration (> 8.0 G, assumingly spec values), which may have created far-traveling surface waves (Franke 1951) that blurred the tactile signal. Even though frequency seems to have a small effect on spatial acuity (Cholewiak et al. 2001; Cholewiak and McGrath 2006), this probably does not apply here as the frequency tested in these studies ranged from 80 to 250 Hz. Another reason for the low accuracy in Novich and Eagleman (2015) might be the paradigm. They asked participants to choose whether they perceived one or two stimuli, even though “one” was never presented. This may have led participants to choose “one” because they expected “one” to be a required answer at some point which would underestimate the accuracy for the N ERMs.
Tactile acuity was higher for horizontal (medial–lateral) than vertical presentation (proximal–distal). Across all conditions, participants performed better when differentiating between columns than rows. Similar and possibly related tactile anisotropies have been found for pressure stimuli in various settings, for instance, for gap detection tasks (Gibson and Craig 2005), absolute localization (Margolis and Longo 2015; Medina et al. 2018), or when participants judged inter-stimulus distances (Longo and Haggard 2011). For vibrotactile stimulation, however, the results are mixed, with some studies finding anisotropies (Sofia and Jones 2013) and others not (Van Erp 2005). This vibrotactile anisotropy has implications both for tactile acuity measurements and for designing tactile displays.
According to the results of Gibson and Craig (2005), the direction and degree of anisotropy is inconsistent across locations suggesting influences of a complex network of variables. Liang and Boppart (2010) quantified the viscoelastic properties of human skin, testing orientations parallel or orthogonal to the Langer’s lines (topological lines corresponding to the natural orientation of collagen fibers in the dermis; Langer 1978) and reported that skin stiffness is anisotropic, depending on the orientation of Langer’s lines. Skin stiffness is more parallel to the Langer’s lines than in the orthogonal direction. The surface wave caused by vibrating stimuli could be more strongly inhibited along the Langer’s lines, facilitating differentiation between two vibrating stimuli. Given that Langer’s lines in the lower thoracic region run medial–lateral, differentiating between columns in a tactile display (ventral–lateral stimulation) should be more accurate than differentiating between rows (dorsal–proximal). However, Liang and Boppart (2010) only found this for high frequencies (600 Hz), while for frequencies of 50 Hz, measurements of skin stiffness in both directions were comparable (as in Sofia and Jones 2013). Even though skin anisotropy may partly be related to stimulus orientation with respect to Langer’s lines and, in the case of hands, to skin ridges (Vega-Bermudez and Johnson 2004; Wheat and Goodwin 2000), other mechanisms appear to be involved.
It has been suggested that the receptive fields of primary afferents and their higher-order neurons may be oval shaped and elongated along the proximal–distal axis (Stevens and Patterson 1995; Cody et al. 2008). Even though there is no evidence for distortions in the shape of the receptive fields of afferent fibers, there are anisotropies in the shape of receptive fields of neurons in the spinal cord and somatosensory cortex (Brown et al. 1975; Alloway et al. 1989). Medina et al. (2018) suggested that the directional bias commonly found in absolute localization tasks for touch (Margolis and Longo 2015) may reflect distortions of a supramodal representation of the skin surface and demonstrated that the directional bias can be modulated by gaze direction. Additionally, attentional mechanisms and the enhancement of resolution at anchor points (joints, spine, see discussion below) have been suggested as possible variables modulating tactile anisotropy (Cody et al. 2008; Medina et al. 2018).
The spine as anchor point
Overall, localization accuracy was lower in the spine area than more peripherally. Vibrotactile stimuli directly located at/or crossing the body midline were more poorly localized than stimuli along the spine. This effect only involved the horizontal presentation direction, however, which may reflect a floor effect due to the lower overall vertical accuracy. These results contradict the common finding of increased tactile acuity with closer distance to anchor points (Boring 1942; Cody et al. 2008; Cholewiak and Collins 2003; Cholewiak et al. 2004). It is worth noting that although both body midline and limb landmarks are usually subsumed under the term of anchor points, the results of studies on limb areas (Boring 1942; Cody et al. 2008; Cholewiak and Collins 2003) might not be directly applicable to the body midline. Wrist and elbow are often referred to as points of mobility (Boring 1942), and Cody et al. (2008) have argued that increased tactile acuity may contribute to improved proprioceptive guidance of active wrist movements. The spine cannot serve the same function and, although there is evidence for a similar effect of higher acuity for the body midline, other neurocognitive mechanisms might underlie this finding (Cholewiak et al. 2004).
A probable explanation is the increased spread of vibration along the dorsal vertebra of the backbone, a key difference between vibrotactile and tactile studies. The characteristics of surface waves spreading from a vibrating source depend strongly on the physical properties of the skin and its underlying tissue (Boyer et al. 2007; Liang and Boppart 2010). We ensured that the tactors were firmly pressed against the lower thoracic area. Due to the lack of underlying damping tissue between the tactors and dorsal vertebra, vibrotactile stimulation probably spread further beyond the tactors that were located directly at the backbone than alongside of it. Cholewiak et al. (2004) found higher localization accuracy for vibratory stimuli at the spine, which appears to contradict our findings. But note that they used substantially bigger tactors and much higher inter-tactor distances (at least 64 mm), with one of their tactors located at the spine, covering the whole dorsal vertebra. Here, three tactors (8 mm) were placed within the same area, and differences between 10 mm inter-tactor distances were reported. This increased sensitivity allowed for more fine-grained assessment and may therefore yield different results. Further studies will have to explore the detailed characteristic of the localization distortion, for instance by precisely mapping the mislocalization errors in the spine area, and find ways of attenuating the spread of the vibration.
It is important to emphasize that the vibrotactile discrimination accuracy rates are limited to the lower thoracic region. Tactile spatial acuity differs greatly by body location due to variations in mechanoreceptor density, which is higher on glabrous than hairy skin (Bolanowski et al. 1994), and lower spatial acuity of passive (e.g., torso, arms and legs) than active body areas (Weinstein 1968). We focused on spatial acuity of the lower thoracic region since such passive areas are better suited to tactile presentation than active parts like the tongue, feet and hands, as they need to be available for performing other functions (Kristjánsson et al. 2016). The lower tactile resolution of passive areas like the torso can be compensated for by the larger skin area that can be stimulated.
The reported tactor acceleration reflects information from the manufacturer and only applies to operation without load (referred to as spec values). When tactors are pressed against the skin (e.g., with straps or elastic fabric, as in our experiment and as common for tactile applications), their characteristics change. Hence, the load frequency for each tactor type was assessed specifically for the experimental setup, showing that, when being exposed to the same pressure, the frequency of both ERM tactor types decreased by approx. 100 Hz, while the frequency of LRAs increased by 28 Hz. Compressing the LRAs under load modifies the resting position of the internal spring, which leads them to vibrate at higher frequency. Azadi and Jones (2014) further found a higher resonant frequency when the LRAs were placed on the finger as compared to the forearm, indicating that their resonant frequency depends on the stiffness of the skin they are mounted on. This notable difference between spec and load values, as well as the inconsistency in their change under load condition, emphasizes the importance of reporting load values additionally to spec values. So far, only a few psychophysical studies on vibrotactile spatial acuity involving tactors have reported load characteristics (Azadi and Jones 2014; Cholewiak et al. 2004; Sofia and Jones 2013). Further considerations regarding load values are that there is no standardized way of measuring them, and that load assessment is not feasible in experimental setups with a closed apparatus design (when the tactors are encompassed by a tactile device). Note also that the actual load exerted on each individual tactor can vary across participants and even within the same participant, depending on physique, posture and breathing. As descriptions of apparatus often lack sufficient detail for replication, we recommend discussion of the spec characteristics, supplementing them with available load values and establishing a standardized way of measuring load characteristics to ensure comparability.
Furthermore, the ratio of tactors and area varies across studies. Eskildsen et al. (1969) tested 5 × 1 tactor arrays, van Erp (2005) 14 × 1 and 11 × 1 arrays and van Erp et al. (2005) tested 8 tactors. Lindeman and Yanagida (2003) measured absolute acuity with an array of 3 × 3 tactors with 60 mm spacing finding an accuracy of 84%. Jones and Ray (2008) used an array of 4 × 4 tactors with the same spacing finding an average accuracy across all tactors of 59%. Although the number of tactors differs considerably between these studies, accuracy by distance was similar. Cholewiak et al. (2004) found no consistent effects of tactor number on localization, concluding that the most important factor for localization accuracy is the inter-tactor distance. In line with these findings, the results of Jóhannesson et al. (2017) suggest that decreasing the size of the area of vibrotactile stimulation does not significantly affect the thresholds for relative vibrotactile spatial acuity.
Additionally, results acquired with the relative point localization (rPL) method, as used here, are not directly comparable to other measurement methods, like absolute point localization (aPL, Sofia and Jones 2013) and two-point thresholds (2PT; Weber 1834). Weinstein (1968) found that spatial tactile acuity with the 2PT was two to four times lower than with the aPL, although they were highly correlated. However, the 2PT method cannot be directly applied to vibrating stimuli, since decisions whether one or two tactors are activated can be affected by additive tactor intensity. As discussed above, Novich and Eagleman (2015) introduced a fake stimulation condition to avoid additive intensities when applying the 2PT method with vibrating stimuli, which might lead to an underestimation of spatial acuity. Even though many studies have used aPL (Cholewiak and McGrath 2006; Lindeman and Yanagida 2003; Sofia and Jones 2013), the ability to localize a point of vibrotactile stimulation may not accurately reflect relative spatial acuity (Jones 2011).
All participants were young adults aged from 20 to 26 years, so generalization to older groups requires caution, since vibrotactile acuity decreases with age, especially for high frequencies (Deshpande et al. 2008). Stevens and Patterson (1995) gathered 1478 individual tactile acuity thresholds, finding that tactile acuity decreases by approximately 1% annually. Devices aimed at helping the elderly should therefore be designed with the caveat that we may be overestimating vibrotactile acuity.