Probing mental representations of space through sketch mapping: a scoping review

The paper addresses how people externalize mental representations of environmental-scale spaces through sketch mapping to understand spatial cognition and navigation. Everyday navigation requires decisions about routes, distances, and landmarks, supported by cognitive maps—internal representations comprising object locations and paths. Prior work shows these representations are distorted relative to geometric properties (e.g., distance, angle) and vary across individuals, influencing environmental learning and memory. Accurate spatial knowledge supports tasks from navigating complex buildings to piloting aircraft. Perception (including visual scene processing), attention, and memory are central for integrating and storing spatial information; neuroscientific evidence implicates the retrosplenial cortex in encoding spatial memory traces. Differences exist between horizontal and vertical spatial processing, with greater distortions in vertical judgments. Sketch maps provide a means to externalize cognitive maps but rely on higher-order processes (e.g., working memory, problem solving). Traditional 2D sketching can inadequately represent 3D mental models (e.g., multi-floor buildings), introducing distortions. This review focuses on environmental-scale spaces (per Montello’s classification), where knowledge is integrated over movement and time, to synthesize methods and cognitive processes probed via sketch mapping and to identify research gaps and opportunities.

Prior literature establishes that cognitive maps guide navigation but are systematically distorted (e.g., Tversky). Individual differences affect acquisition of metric knowledge and integration of separately learned places (Ishikawa & Montello). Visual scene processing and attention underpin the formation of spatial memory; the retrosplenial cortex stores experience-dependent spatial traces. Research differentiates horizontal and vertical spatial perception, with vertical judgments prone to larger distortions. Sketch maps have been used for decades to visualize mental representations, though their generalized and incomplete nature complicates metric evaluation. Conventional pen-and-paper 2D sketching predominates, but translating 3D environments to 2D surfaces can cause scaling and information loss, particularly for multi-level buildings. Emerging VR/AR sketching tools have potential to better capture vertical relationships. The review situates itself within this literature by focusing on environmental-scale spaces, where spatial memory differs from vista-scale memory due to path length and sequential learning effects.

The scoping review followed PRISMA-ScR guidelines. A protocol was prepared and approved by authors prior to the search (not registered). Information sources: PubMed, Web of Science Core Collection, Scopus, Dimensions, and Embase. Inclusion criteria: English-language articles with adult participants, published up to December 15, 2023; systematic reviews included; environments restricted to Montello’s environmental scale; peer review not required (four non-peer-reviewed conference papers assessed for quality). Search strategy combined terms: ("cognitive map" OR "spatial representation" OR "mental model" OR "cognitive collage" OR "mental representation" OR "navigation" OR "wayfinding" OR "spatial memor*" OR "cognitive process") AND "sketch*". Selection procedure: three-stage screening (titles, abstracts, full texts) by two independent raters, with discrepancies resolved by a third rater. Raters underwent training with sample sets prior to screening/assessment to ensure consistency. Data abstraction: a standardized form was developed a priori; two raters extracted data independently and reconciled differences with a third rater. Data were organized into tables (Appendices 1–2) with taxonomy informed by Hegarty et al. (2006) and Krukar et al. (2023). For three papers without explicit space acquisition (design tasks), the space acquisition phase was labeled "Design task" per original authors. Cognitive processes identification protocol: Using Harvey’s (2019) domains, the team keyword-searched Methods/Results/Discussion sections for terms related to sensation, perception, motor skills/ability, attention, memory/recall, executive function/reasoning/problem-solving, and processing speed (excluding Introductions). Occurrences were tallied to quantify how many papers addressed each domain (Fig. 2). Statistical analysis: A contingency table examined the relationship between space acquisition mode (past experience, direct experience, virtual, design, videos/maps) and sketching medium (pen/blank paper, pen/paper with information, digital, 3D VR). Given low cell counts, Fisher’s exact test (R 2023.06.2+561) tested independence; conceptually similar categories were merged to ensure validity. Synthesis: Descriptive synthesis summarized methods and tasks, cognitive processes, and sketch map assessments, and compiled reported correlations between sketch-map metrics and other spatial knowledge measures (Table 3). PRISMA flow: 1866 records identified; 923 duplicates removed; 943 titles screened; 99 abstracts screened; 39 full texts assessed; 22 included; plus 2 added via co-authors/cross-references, totaling 24.

• Study corpus and characteristics:
  • 24 studies (1970–2023); 15 journal articles, 9 conference papers. Sample sizes ranged from 6 to 460; 17 studies reported detailed demographics.
  • Sketching media: pen-and-paper (n=18), digital (n=4), hybrid pen-and-paper + digital (n=1), VR digital sketching (n=1).
  • Environments: 19 studies without vertical components (e.g., neighborhoods), 5 with vertical components (e.g., buildings).
• Cognitive domains identified (Harvey, 2019) across included studies (Fig. 2 counts):
  • Memory: 18 papers; Perception: 9; Executive functioning: 5; Attention: 4; Motor ability: 1; Sensation: 1; Processing speed: 0; Language/verbal skills: 0.
• Methodologies and assessments (Fig. 3):
  • Space acquisition modes (% of studies): Past experience 43%; Design 19%; Virtual environment 19%; Direct experience of unfamiliar environment 17%; Videos or maps 0%.
  • Sketching medium (%): Pen and paper (both information) 50%; Computer/tablet 19%; Pen and paper (some information) 17%; 3D 4%; VR 4%; Other 6%.
  • Sketch map assessment types (%): Quantitative 49%; Qualitative 21%; Both 17%; No assessment 13%.
• Contingency between space acquisition and medium (Table 2; counts):
  • Direct experience: Digital 1, Pen/paper 4 (Total 5)
  • Past experience: Digital 3, Pen/paper 8 (Total 11)
  • Virtual: Digital 1, Pen/paper 5 (Total 6)
  • Design: Digital 0, Pen/paper 3 (Total 3)
  • Totals: Digital 5, Pen/paper 20 (Total 25 combinations)
  • Fisher’s exact test: no significant association (p = 1), indicating independence of acquisition mode and sketching medium.
• Correlations between sketch map metrics and spatial knowledge (Table 3 highlights):
  • Billinghurst & Weghorst (1995): Map goodness correlated with knowing where everything is (Virtual Valley r=0.635, p<0.05; Neighborhood r=0.405, p<0.05; Cloudlands ns) and with orientation (Virtual Valley r=0.738, p<0.05; Neighborhood r=0.524, p<0.05; Cloudlands ns). Map goodness ranking correlated with ease of interaction, navigation, movement, and tool use in Virtual Valley (rs 0.645–0.882, p<0.05; other worlds ns).
  • Dong et al. (2022): Sketch map performance showed a weak negative correlation with direction/distance errors and a strong positive correlation with distance estimation accuracy (see their Fig. 8C for full Pearson values).
  • Gehrke et al. (2018): Sketch map usefulness correlated with perspective taking (r=0.39, p<0.05) and experienced realism (r=0.41, p<0.05).
  • Jaeger et al. (2023): More target details associated with better pointing (within r=-0.35, p=0.01; between r=-0.29, p=0.04). Route details score correlated with Silcton performance (r=-0.37, p<0.01), between-pointing error (r=-0.52, p<0.001), and model building (r=0.37, p<0.01).
  • Wang & Schwering (2015): Sketch map accuracy not correlated with sketching/artistic/cartographic skill or geoscience background; greater familiarity yielded more complete and accurate maps (qualitative statement, no inferential stats provided).
• Synthesis insights:
  • Quantitative assessments often emphasize metric accuracy, potentially undervaluing the generalized/incomplete nature of sketch maps. Qualitative spatial relations offer a bridge by enabling objective scoring without reliance on exact Euclidean metrics.
  • Three design-focused studies did not evaluate sketch performance, emphasizing the sketching process over outcomes.
  • No consensus emerged on matching acquisition modes to sketching media; VR/AR sketching shows promise, especially for vertical components, and presence/immersion may covary with sketch accuracy.

The review reveals methodological diversity and an inherent tension: researchers often collect sketch maps in unstructured, naturalistic settings yet analyze them with predominantly quantitative, metric-focused methods. This can misrepresent the value of generalized spatial knowledge that sufficiently supports behavior. Formal analyses of qualitative spatial relations can bridge qualitative assessments and quantitative rigor. Correlations between sketch map metrics and independent spatial knowledge measures are observed across studies, suggesting sketch maps are valid indicators of environmental knowledge and navigational performance; however, heterogeneity in metrics and validation tasks prevents generalizable conclusions about when sketch maps are most valid (for which environments/users). The analysis found no association between space acquisition modes and sketching media, indicating a lack of consensus on methodological pairing; this invites targeted research to derive recommendations. Cognitive processes emphasized include memory and perception, while attention, executive functions (reasoning/problem solving), and motor aspects are underrepresented. Notably, many studies do not distinguish short-term from long-term memory in sketching tasks, limiting interpretability. Emerging VR/AR sketching tools could better capture vertical spatial relations and may benefit from measuring presence/immersion; objective process measures (e.g., EEG, eye tracking) could quantify cognitive load and dynamics during sketching.

This scoping review of 24 studies provides a comprehensive overview of how cognitive maps of environmental-scale spaces are externalized via sketch mapping. It documents substantial methodological diversity across acquisition tasks, sketching media, and assessment approaches, and identifies memory and perception as the most frequently targeted cognitive processes. The work underscores a methodological conflict between unstructured, ecologically valid sketching protocols and preference for structured quantitative evaluation, and highlights underused media (e.g., digital, VR/AR) that may be better suited to certain spatial contexts, especially with vertical components. Future research should specify targeted cognitive processes (e.g., differentiate short-term vs. long-term memory), converge on established analysis frameworks (including qualitative spatial relations), systematically investigate medium–acquisition pairings, and incorporate objective measures (e.g., neural, eye tracking) to deepen understanding of the sketching process and improve the design of navigational aids and spatial representations.

• Heterogeneity of methods (tasks, media, and assessment metrics) limits cross-study comparability, replicability, and evidence-based recommendations.
• Underrepresentation of environments with vertical components constrains generalizability to multi-level spaces.
• Many studies treat sketch maps as a coarse index of memory without distinguishing short-term from long-term memory, reducing interpretive precision regarding cognitive processes.
• Quantitative assessments often prioritize metric accuracy, potentially biasing evaluation against the generalized nature of sketch maps.
• Protocol was not registered; inclusion restricted to English-language, adult-participant studies; some included studies were non–peer-reviewed conference papers (though assessed for quality), which may introduce bias.
• Limited use of objective process measures (e.g., neural or ocular metrics) to assess cognitive load and dynamics during sketching.