"I can't name it, but I can perceive it" Conceptual and Operational Design of "Tactile Accuracy" Assisting Tactile Image Cognition
Jiangtao Gong, Wenyuan Yu, Long Ni, Yang Jiao, Ye Liu, Xiaolan Fu, Yingqing Xu · 2020 · Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2020) · doi:10.1145/3373625.3417015
Summary
This paper introduces "tactile accuracy," a new evaluation criterion for measuring how well blind people recognize objects in raised-line tactile images, replacing the conventional "naming accuracy" borrowed from visual image recognition research. The fundamental problem is that touch and vision operate very differently: the tactile perception field is much smaller than the visual field, tactile sensation processes information serially rather than in parallel, and it demands more working memory. Consequently, a person exploring a tactile image by touch may correctly perceive the shape of an object but be unable to name it — for example, recognizing a "jellyfish" shape but calling it a "spider plant" due to similar tactile features. Using naming accuracy alone (which averaged only 12.78% across all subjects) would falsely conclude that these subjects extracted no useful information. The researchers designed a multi-level coding scheme that evaluates recognition at progressively broader levels: exact naming, third-level categorization (e.g., mammals), second-level (e.g., animals), first-level (natural vs. man-made), identification of similar objects from different categories, and other samples from the same category. "Tactile accuracy" is defined as positive when any of these levels indicates successful shape information extraction. The study tested 30 Chinese university students (10 blindfolded sighted, 10 congenitally blind, 10 late blind; ages 20-27) on 242 raised-line images of common objects printed on a Braille printer, with each image explored by touch for up to 30 seconds.
Key findings
Overall naming accuracy was only 12.78% (SD=9.19%), but tactile accuracy reached 29.52% (SD=14.76%), and first-level categorization accuracy was 65.40% (SD=12.48%) — revealing that subjects extracted far more shape information than naming alone would suggest. Critical differences emerged between subject groups: later blind subjects had significantly higher naming accuracy (13.84%) and tactile accuracy (36.03%) than congenitally blind subjects (8.84% naming, 19.88% tactile), confirming that prior visual experience aids tactile recognition. Blindfolded sighted subjects achieved the highest naming accuracy (15.67%) but their tactile accuracy (32.60%) was lower than later blind subjects, suggesting blind people develop superior tactile shape perception even if they struggle to link shapes to names. Multiple regression analysis identified key design factors: perspective (3D viewpoint) was the strongest negative predictor of tactile accuracy — none of the 30 subjects recognized a bed or nut shown in 3D perspective. Image symmetry was the strongest positive predictor: a butterfly (symmetric) was recognized by 28/30 subjects while asymmetric pigeons and eagles were recognized by nearly none. Object rigidity positively predicted recognition (rigid objects are easier to recognize by touch). Omissive lines for overlap (where foreground objects occlude background, breaking contour lines) negatively predicted recognition. Image size relative to actual object size also mattered — both oversized and undersized images impaired recognition.
Relevance
This paper makes a fundamental methodological contribution to tactile graphics research by demonstrating that the standard evaluation metric (naming accuracy) systematically underestimates blind users' perception of tactile images. The "tactile accuracy" criterion better captures what touch actually conveys and can guide both image design and personalized learning strategies. For accessibility practitioners creating tactile materials, the four concrete design guidelines are immediately actionable: (1) avoid 3D perspective — use parallel or cut-slice views instead; (2) preserve main contours without occlusion breaks; (3) match image size to actual object proportions; (4) preserve symmetry in symmetric objects. The finding that different blind populations have different cognitive profiles for tactile recognition — congenitally blind subjects benefit most from training on object rigidity concepts, later blind subjects need familiarity with image size ratios, and newly blind adults should attend to overlapping line patterns — has direct implications for personalized tactile literacy education. Limitations include the all-Chinese undergraduate sample (cultural familiarity with objects may differ), the preliminary operational definition of tactile accuracy, and the use of a single image set.
Tags: visual accessibility · tactile graphics · blindness and low vision · tactile accessibility · cognition · perception · education · design methodology