Viewing Angle and Task Materials Shape Visual Perspective Taking in Preschool Children

Four-year-old children judging how objects appear from another person’s point of view perform significantly better when viewing scenes from directly opposite positions than from right angles, according to an experimental investigation published in Frontiers in Psychology. The study, conducted by researchers at Hainan Normal University in Haikou, China, evaluated level-2 visual perspective taking among preschool children using a standardized picture-selection paradigm. Level-2 visual perspective taking represents the social-cognitive capacity to understand that even when two observers look at the same physical display, their visual experiences and the apparent spatial configurations of objects differ based on their respective viewing positions. By systematically manipulating both the angular disparity between observer viewpoints and the physical characteristics of the test materials, the investigators demonstrated that perspective-taking accuracy does not decline monotonically as viewing angles widen. Instead, accuracy was highest when viewpoints coincided at zero degrees, fell to intermediate levels at one hundred eighty degrees, and dropped to its lowest point at orthogonal ninety-degree angles.

The developmental framework underpinning the investigation traces back to foundational developmental psychology, specifically the classic three mountains task introduced by Jean Piaget and Bärbel Inhelder in 1956. While early spatial perspective research established that young children frequently experience difficulty moving beyond their own egocentric viewpoints, subsequent theoretical formulations refined the construct into distinct developmental levels. In 1981, John Flavell and colleagues established a formal distinction between level-1 and level-2 visual perspective taking. Level-1 perspective taking requires children to make simple visibility judgments, determining whether an object falls within another person’s line of sight or is occluded by obstacles. In contrast, level-2 perspective taking demands that children mentally reconstruct how an object or scene appears from another vantage point, inferring specific spatial orientations, front-to-back relations, or relative positions. Because level-2 processing involves mentally representing an alternative viewpoint while simultaneously inhibiting one’s own immediate perspective, it imposes substantial social-cognitive and executive demands that typically show protracted development across early childhood.

To explore how visual perspective judgments operate under controlled conditions, the researchers recruited a sample of 163 four-year-old children from public kindergartens across urban and township areas in four administrative regions of Hainan Province, comprising Haikou, Danzhou, Wanning, and Qiongzhong Li and Miao Autonomous County. The initial recruitment enrolled 190 children through stratified random sampling, after which 27 participants were excluded for failing to meet the strict age eligibility criteria of being between 48 and 59 months old. The final cohort consisted of 80 boys and 83 girls with a mean age of four years and four months and a standard deviation of 3.2 months. All enrolled children had no documented history of neurodevelopmental disorders or visual impairment, understood spoken Mandarin instructions, and had written informed consent provided by their legal guardians. An a priori statistical power analysis indicated that a minimum sample size of 128 participants would provide eighty percent power to detect medium effect sizes within a repeated-measures experimental design, confirming the study was adequately powered.

The experimental procedure employed a picture-taking paradigm originally adapted from developmental protocols designed by Andrea Frick and colleagues. This setup framed the cognitive assessment as a photographer game, allowing children to indicate another observer’s viewpoint through nonverbal picture selection rather than demanding complex verbal descriptions. Throughout testing, each child remained seated in a fixed position facing a tabletop display, while the angular position of a toy photographer figurine was manipulated across three distinct rotation angles: zero degrees, where the photographer stood beside the child sharing the exact same viewing direction; ninety degrees, where the photographer stood orthogonal to the child on the left or right; and one hundred eighty degrees, where the photographer stood directly opposite the child facing them face to face. In addition to rotation angle, the experiment manipulated test materials across two distinct conditions: a Lego condition featuring standard four-centimeter-tall human-like figurines with identifiable facial features, limbs, and bodily postures, and a geometric condition featuring combinations of solid wooden blocks, including cones, cylinders, cubes, and rectangular prisms.

The formal assessment utilized a fully within-subjects two-by-three factorial design, crossing the two test tools with the three rotation angles. Each of the six experimental conditions comprised three distinct four-alternative forced-choice trials, yielding a total of eighteen formal test trials per participant. Across all trials, children were presented with four candidate photographs arranged in a two-by-two array on an A4 sheet, depicting the display from front, back, left, and right vantage points in randomized positions to eliminate spatial location biases. In the Lego figurine condition, children judged which photograph captured the orientation and visual appearance of a single figurine. In the geometric condition, two solid objects were placed together on the table alongside the photographer, requiring children to determine the relative spatial layout, such as left-right and front-back arrangements, from the photographer’s perspective. Experimenters recorded children’s unambiguous pointing responses against a preset objective answer key, awarding one point for each accurate selection and zero points for incorrect choices, resulting in condition subscores ranging from zero to three.

Statistical evaluation through a two-way repeated-measures analysis of variance revealed significant main effects for both rotation angle and test tool. The main effect of rotation angle was highly significant, yielding an F-ratio of 72.383 with a partial eta squared of 0.309 after applying Greenhouse-Geisser corrections for sphericity violations. Post hoc pairwise comparisons with Bonferroni corrections confirmed significant differences among all three viewing angles. The zero-degree baseline, which required no mental transformation, produced the highest mean score of 1.92 out of 3, followed by the one hundred eighty-degree condition with a mean of 1.42, while the orthogonal ninety-degree condition yielded the lowest mean score of 1.08. This nonmonotonic performance profile demonstrated that cognitive difficulty in perspective transformation does not scale linearly with increasing angular displacement. Separate one-way analyses confirmed that this nonmonotonic angle pattern was robust across both the Lego figurine task and the geometric solid task, although the difference between ninety and one hundred eighty degrees in the geometric task alone was statistically marginal.

The analysis of variance also demonstrated a substantial main effect of test tool, yielding an F-ratio of 125.279 and a large effect size with a partial eta squared of 0.436. Children achieved markedly higher raw correct scores when evaluating Lego figurines, averaging 1.77 points across trials, compared to an average of 1.17 points when evaluating geometric solids. Descriptive condition means illustrated this consistent performance gap: in the zero-degree alignment, children scored an average of 2.18 on Lego trials versus 1.66 on geometric trials; at ninety degrees, scores dropped to 1.34 for Lego versus 0.82 for geometric; and at one hundred eighty degrees, scores reached 1.80 for Lego compared to 1.04 for geometric solids. Despite this uniform performance advantage for the figurine materials, the interaction between test tool and rotation angle did not reach statistical significance, producing an F-ratio of 2.397 and a p-value of 0.093, suggesting that the relative impact of viewing angle operated in a broadly comparable manner across both types of visual materials.

Beyond overall accuracy, the researchers conducted detailed analyses of egocentric errors, specifically examining whether children erroneously selected the photograph matching their own immediate viewpoint when perspective transformation was required. In the non-zero rotation trials, the photograph representing the child’s personal view served as an egocentric distractor. Across 489 total presentations per condition, egocentric errors occurred in 28.83 percent of Lego ninety-degree trials, 31.70 percent of Lego one hundred eighty-degree trials, and 23.31 percent of geometric ninety-degree trials. However, egocentric errors surged dramatically in the geometric one hundred eighty-degree condition, where they accounted for 50.10 percent of all responses, with 80.98 percent of children committing at least one egocentric error. A two-by-two repeated-measures analysis of variance on error proportions confirmed a significant main effect of tool, a significant main effect of angle, and a highly significant tool-by-angle interaction with an F-ratio of 34.794. Bonferroni-corrected simple-effects testing revealed that egocentric choices were significantly more prevalent at one hundred eighty degrees than ninety degrees within the geometric condition, whereas error rates between ninety and one hundred eighty degrees did not differ significantly in the Lego condition.

Supplementary analyses evaluated the potential influence of participant sex by incorporating sex as a between-subject factor in a mixed-design three-way analysis of variance. This analysis revealed no significant main effect of sex, yielding an F-ratio of 0.674 and a p-value of 0.413. Furthermore, none of the interaction terms involving participant sex approached statistical significance, including the tool-by-sex interaction, the angle-by-sex interaction, and the three-way tool-by-angle-by-sex interaction. Correlation analyses between children’s total performance on the Lego task and their total performance on the geometric task revealed only a weak positive association, with a Pearson correlation coefficient of r = 0.244 and an r-squared value of approximately 0.06. This weak correlation indicates that the two assessment tasks shared only six percent of common variance, suggesting that they may engage distinct cognitive operations rather than simply reflecting alternate measurements of a single uniform psychological construct.

The study authors emphasize several critical methodological and interpretive limitations regarding their findings. First, the experimental design did not measure processing speed, response latencies, eye movements, or explicit strategy reports, meaning the observed differences represent behavioral performance outcomes rather than direct measurements of underlying cognitive mechanisms or mental self-rotation. Second, the comparison between Lego figurines and geometric solids contained multiple confounded dimensions, including task structure, visual discriminability, and social cues. The Lego condition evaluated orientation judgments of a single object possessing salient facial and postural features, whereas the geometric condition required spatial-relational reasoning across two objects, several of which had rotationally or contour-symmetric geometries. Consequently, the performance advantage observed for figurines cannot be attributed purely to social cues without future studies matching task structures. Additionally, because the experimental design tested only three discrete angles, the findings document behavioral differences among specific viewpoints rather than establishing a continuous mathematical curve of perspective transformation across early childhood development.

Subject of Research: Effects of viewing angle and task materials on level-2 visual perspective taking in four-year-old children

Article Title: What factors influence level-2 visual perspective-taking ability in 4-year-old children?

Article References: Ding, J., Chen, J., Bao, J., Lin, J., Jia, L., Hu, P., Liu, X., Jiang, J., & Diao, J. (2026). What factors influence level-2 visual perspective-taking ability in 4-year-old children?. Frontiers in Psychology, 17, Article 1867009. https://doi.org/10.3389/fpsyg.2026.1867009

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1867009

Keywords: visual perspective taking, cognitive development, preschool children, spatial cognition, theory of mind, egocentric errors, developmental psychology, viewpoint transformation, spatial orientation, childhood cognition, experimental psychology, picture taking task, mental representation, social cognition, spatial reasoning, inhibitory control, rotation angle, stimulus discriminability