Canonical Size in Haptic Drawings

Canonical size—a term coined by Konkle and Oliva (2011)—concerns to how mental representations of object represent size information. Abovementioned authors treat this phenomenon as a matter of visual cognition (see also Konkle & Oliva, 2012) and provide evidence on its existence in a series of studies on visual modality, including an experiment regarding drawing from memory. In their view, canonical size specifies the optimal distance of the observer in the environment from a viewed object. In other words, representations of familiar objects depend on the memorised real-world physical size of a particular object in relation to a frame of space around the object concerned.

Cognitive psychologists working on the representation of objects assume the existence of multimodal (Lacey et al., 2007; Lacey & Sathian, 2012) or amodal (Loomis et al., 2013) mental representations which are flexibly accessible via perceptual (e.g., visual and haptic) and memory inputs and are spatial in format. Spatial imagery contains information about the simplified object form and its size. Importantly, this size information is coded in relation to other objects’ sizes (Loomis et al., 2013). Therefore, in our opinion, canonical size may be revealed in a task performed in the haptic domain if “object representations are reactivated from long-term memory representations relative to a space around them”—as stated by Konkle and Oliva (2011, p. 25). Although almost no one (except individuals with visual impairments) does it daily this way, drawings can be created under haptic control using materials designed for people with special educational needs. Focusing on such haptic drawings, we have examined the phenomenon of canonical size in the haptic domain. We hypothesised that objects which are physically larger in the world are drawn as larger than objects which have smaller real-world size.

This study included 32 university students who had normal or corrected to normal vision (16 females) aged between 18 and 25 years (M = 21.72, SD = 1.44). A priori power analyses using G-Power 3.1 yielded that 24 participants would be needed to detect both a within-participants and between-participants effect in a mixed model ANOVA, based on an effect size of f = .25 (cf. Konkle & Oliva, 2011).

Participants were randomly assigned to one of two experimental groups. Sixteen participants (eight females) drew with a stylus on special foil for embossed drawings in an A4 format, and sixteen on an A5 format foil (eight females). These foil sheets were arranged horizontally and placed on a special rubber mat. All participants were blindfolded when tested individually. They were instructed to control the drawing process by touch. More specifically, participants were asked to explore by a nondominant hand the convex shape during its creation—when the drawing was in progress. In line with the study from Konkle and Oliva (2011), participants were instructed to draw (in the random order) sixteen objects belonging to eight different ranks of a real-world size category (see Figure 1), each drawing of the named object on a separate sheet. Importantly, any words connected with the size of objects were never used.

OPEN IN VIEWER

Figure 1.

Drawn size (A) and the ratio of drawn and foil sizes (B) as functions of real-world object size rank for different foil sizes. Error bars represent ±1 SEM.

Taking the Konkle and Oliva (2011) approach, we assessed the size of the drawings (in centimetres) by measuring the length of the diagonal of the rectangle in which the drawing was located. This bounding box was set around an object, after ignored extraneous objects (e.g., fence near the house, smoke from the chimney).

An analysis of variance (ANOVA) with the between-group factor of foil size (2), the within-group factor of size rank (8), and the dependent variable of drawn size revealed two statistically significant effects. First, the analyses yielded effect of the foil size, F(1, 30) = 9.11, p = .005, η p 2  = .23, as drawings made in the A4 foil were larger (M = 10.71, SE = 0.65) than drawings made in the A5 foil (M = 7.95, SE = 0.65). Second, the main effect of size rank was significant, F(4.69, 140.79) = 45.28, p < .001, η p 2  = .6, which was best described as a linear function between size rank and drawn size, F(1, 30) = 132.83, p < .001, η p 2  = .82 (see Figure 1A). An analogous ANOVA with the dependent variable of a ratio of the object to the frame across foils sizes (i.e., length of the diagonal of the rectangle in which the drawing was located divided by length of the diagonal of the foil) yielded only the significant effect of size rank, F(4.78, 143.43) = 46.50, p < .001, η p 2  = .61. This effect was best described as a linear function between size rank and the ratio of the drawn object and foil diagonal length, F(1, 30) = 146.07, p < .001, η p 2  = .83 (see Figure 1B). This time the main effect of size rank did not reach significance, F(1, 30) = 0.3, p = .590.

Our study conducted in the haptic domain replicated the effects found in the task of drawing from memory in the visual domain (Konkle & Oliva, 2011). Namely, we showed a systematic relationship between physical size of familiar objects and drawn objects size. In line with our prediction, participants drew as larger these familiar objects which have larger real-world sizes. Furthermore, the ratio of the sizes of particular drawn object to the frame across foil sizes was consistent. This pattern of results suggests that canonical size may be observed in haptic drawings—made without sight control.