Introduction

The canonical size effect

Research on the canonical size phenomenon originates in classical research on canonical perspective1 which showed that people represent preferred viewpoints of real-world objects. Evidence for similar canonical size, i.e., mentally representing the preferred size of particular objects, which increases with increasing the actual size of objects, originated from a series of experiments by Konkle and Oliva2 on drawing from long-term memory, imagery, and perceptual preference tasks. They coined the term “canonical visual size”, emphasising that the source of this phenomenon is our everyday visual experience of seeing given objects having particular angular (retinal) sizes. In a similar vein, the canonical real-world distance of real-world objects from the observer was recently explained as being a visual phenomenon3.

The original drawing study testing the canonical size phenomenon among adults2 involved creating drawings on paper under visual control based on representations of objects stored in long-term memory. Participants were simply asked to draw a picture of the named object. These objects varied on the real-world size dimension (the size rank was used as an independent variable; notably, the relationship between the actual physical size of these objects and their size rank is logarithmic2), but participants were unaware that the size issue was crucial in the study. Besides, the frame of space in which the object was drawn varied (the paper size was used as an independent variable). The findings revealed that the drawings were larger for larger real-world objects, whereas the drawn size was adjusted to the frame size. In other words, for all paper sheet sizes used, there was a linear increase in drawn size as the size rank of the real-world objects increased. A remarkably similar pattern of results was obtained when conducting this drawing task under haptic control among blindfolded sighted adults4,5,6.

Further evidence that people represent the size of real-world objects comes from behavioural studies on a wide range of tasks within visual perception among adults2,7,8,9,10 and children aged 3 to 4 years11,12. The cerebral basis of this representation was identified in the occipitotemporal cortex13,14.

Although most research on canonical size has been conducted in the visual domain (e.g., drawings from memory were produced under visual control – participants could see what they were drawing), a recent study among blind participants on drawing from memory has revealed that canonical size also characterises how people without any visual experience represent real-world objects15. Hence, the canonical size phenomenon seems spatial in nature, not visual—as initially thought2.

Mental representation of size: developmental changes

Infants as young as 7-month-olds can discriminate between physically larger and smaller objects when looking at them16. Besides, infants aged 7 to 12 months are sensitive to real-world object size and show familiar size knowledge; however, only when looking at real objects and not when these objects are represented in pictures17. Children seem to represent the typical size of common objects at the age of 2. They are capable of identifying within categories of objects ones that are big versus small perceptually and normatively by comparing them, respectively, to another physically present object (perceived via sight) or memory representation of a given class18. With cognitive development, 3-year-olds looking at illustrations (i.e., 2-dimensional colour pictures, each depicting the toy within the food or animal category) already interpret the features of the depicted objects by relating them to the real-world objects (e.g., toys) that the drawings represent19. Findings from experiments on visual size judgments about pictured objects in the Size-Stroop task11 and the visual search task12showed that 3–4-year-olds activate real-world size information from depicted objects, distinguishing between big versus small real-world objects. However, to our knowledge, there has not yet been a study among children in which the size of real-world objects has been systematically manipulated (as in the classic study of Konkle and Oliva 2 on the canonical size phenomenon) to show how subtleties of size sensitivity develop. We decided to fill this gap in the current study by testing the influence of size rank on drawn size using a drawing from memory task adapted from Konkle and Oliva2. It is worth mentioning that the size of objects represented in children’s drawings is a more complex problem than what is analysed in the present research. Canonical size (measuring when only one object is drawn on a standard sheet of paper) seems to occur developmentally earlier than projective size, representing visually realistic size (measuring when a few objects are drawn on one sheet of paper or when a particular visual perspective for presenting an object is suggested)20. However, this issue is beyond the scope of the current paper since we focus on the canonical size in our research.

As size is a spatial property, the internal representation of an object’s size can be created from both visual and haptic experiences15,21. As Long et al. 11 noted, as children grow, the size of various objects in relation to the size of their bodies (e.g., their hands) changes dramatically. It is, therefore, not surprising that size representation matures during childhood. Gradual maturation also seems to involve correspondence between visual and haptic modalities in processing information about size. The developmental changes in cross-modal, visual-tactile object integration begin in early infancy, and touch seems to be the sensory scaffold on which multisensory perceptual development is constructed22. This integration seems to occur earlier in the development regarding such amodal properties of objects as shape and texture when compared to the size property 22,23. A recent study on spatial scaling showed that the functional equivalence between vision and touch21,24,25 is not fully developed in 6–8-year-old children 26. Moreover, cross-modal integration of size information perceived through visual and haptic exploration develops at school age. Children younger than 8 years do not integrate visual and haptic spatial information in a size discrimination task27, and the cross-modal integration of object information perceived visually or haptically matures until 8–14 years of age28. On the one hand, it is fair to assume that regarding the real-world objects’ size information, vision-based information is more stable during childhood than haptic-based information. Children have the opportunity to observe different objects from different distances and experience the difference between angular size and the actual size of objects in a similar way in the following years of life. At the same time, the children’s body size in relation to the size of given objects changes significantly over the first few years of life (e.g. the same object may appear to be a different size depending on whether a small or large hand grasps it). On the other hand, object size estimates in children (before 8 years of age) rely mainly on haptics rather than vision – even when the haptic information is unreliable 22.

In summary, children may learn about an object’s size through visual and haptic exploration. However, the current study focuses on another issue – revealing knowledge of the canonical size of an object in a drawing2 produced under varied perceptual conditions. To our knowledge, the area of representing object size in children’s drawings produced from memory in varied perceptual domains (with and without visual control) has not been investigated. At the same time, as knowledge of an object’s spatial properties can be both acquired and manifested through visual and tactile domains, investigating the canonical size phenomenon among children in varied perceptual conditions (in other words – under varied possible perceptual feedback/control) may lead to a more robust conclusion about the phenomenon tested from the developmental perspective. Hence, it is worth exploring whether the canonical size effect would be revealed in the same way when investigated in the drawing from memory task2 produced under different perceptual conditions among children of various ages. Notably, studies of producing drawings under haptic control by blindfolded sighted children are not common. Still, this group of participants sometimes serves as a control group for children who are blind in drawing development research29, showing that blindfolded sighted children can deal with drawing without visual control.

The present study

The objective of this study was twofold. First, we aimed to test whether the canonical size phenomenon might be observed in the drawing from memory task among children. Second, we explored whether this effect of representing objects differing in size in the physical world at respective sizes in internal representation might be observed in the drawing from memory task produced without or with visual control by children of different ages.

By crossing the factors of perceptual condition (sighted vs. blindfolded) and material used for drawing (regular white paper vs. special foils for producing embossed drawings), we designed four conditions for producing drawings from memory. These conditions provide a comprehensive experimental design and are characterised by varied possible perceptual feedback when producing drawings: (a) visual control/sighted, paper (typical condition for sighted children’s drawing activity) – visual and proprioceptive feedback is available, (b) visual control, foil – visual, haptic, and proprioceptive feedback is available; visual information is likely to dominate the haptic information, (c) no visual control/blindfolded, paper – proprioceptive feedback is available, (d) no visual control, foil – haptic and proprioceptive feedback is available (for more details, see5).

We tested children aged between 5 and 9 years. This age range was chosen based on previous literature on size representation among children and children’s drawing development and abilities in different domains. The youngest tested participants were 5-year-olds because at this age, children distinguish between big and small real-world objects11, and, at the same time, most children of this age are capable of producing recognisable drawings of ordinary objects 30,31,32 and should also be capable of producing relatively recognisable drawings under haptic control (without visual feedback)29. We decided to test children up to 9 years as, at this age, children should be able to integrate and equivalently use spatial information whose source is sight or touch28. With increasing age, children are expected to produce drawings with more details, but the graphic productions should be sufficiently detailed to be recognisable from the early age of 5 years30.

We hypothesised that the larger the real-world object is (in other words, the larger its size rank), the larger it is drawn by children (H1). Thus, we predicted the canonical size effect in children aged 5–9 years, best explained by a linear function between size rank and drawn size2. Moreover, we explored the robustness of the canonical size effect by analysing it among children of various ages who produced drawings with different perceptual feedback.

Methods

Participants

A total of 72 children aged 5 to 9 years (± 3 months) participated in the present study (33 female, 39 male). We tested three age groups of participants: 5-year-olds (n = 24, 9 female; age in months: minimum [min] = 57, maximum [max] = 63, M = 60.08, SD = 1.86), 7-year-olds (n = 24, 12 female; age in months: min = 81, max = 87, M = 83.79, SD = 1.84), and 9-year-olds (n = 24, 12 female; age in months: min = 106, max = 111, M = 108.38, SD = 2.00). All tested children attended kindergarten or primary school. They had normal or corrected-to-normal vision and did not have any diagnosed disabilities. All participants provided verbal assent prior to study participation, and parents gave written informed consent before. Participants received a pen after completing the study as an incentive. The study was approved by the Ethical Committee of the Institute of Psychology of The John Paul II Catholic University of Lublin and was conducted in accordance with the Declaration of Helsinki.

Design

The current study had a mixed design. As in the original experiment conducted in different perceptual domains among sighted adults 5, we manipulated three within-subjects independent variables: material used (paper, foil), (2) perceptual condition (sighted, blindfolded), and size rank (1 to 8, depending on the drawing topic: key, apple, shoe, backpack, dog, floor lamp, car, and house, respectively – representing objects of increasing size rank in the physical world, see 2). Additionally, a between-subjects independent variable of age groups was included since we tested children at three levels of age: 5-year-olds, 7-year-olds, and 9-year-olds.

A dependent variable in this study was drawn size. Following the approach taken in the previous studies in this field, we operationalised the drawn size (in mm) as the length of the diagonal of the rectangle frame bounding a produced drawing 2,4,5,6,15. All lengths were measured using a procedure of scanning drawings at a fixed resolution, determining the boundaries of each object (importantly, all elements drawn additionally beyond the object of interest, e.g., a garden in front of a house or a carpet on which a floor lamp stands, were ignored) and converting the dimension of the bounding box into millimetres using the known resolution. The bounding box selection was done using Photoshop CS6 software (Abobe Inc., USA), while dwawn size calculation was done using a custom script written in MATLAB 2014b (Mathworks, USA).

Materials

We used A4-sized (i.e., 210 mm high and 297 mm wide) sheets of regular white paper, A4-sized special transparent foils for producing embossed drawings (with a special rubber mat in a set, a so-called Swedish raised-line drawing kit), and sharpened pencils as materials.

Procedure

We used the procedure previously used in the study among adults 5, which was modified concerning the original procedure of Konkle and Oliva 2. Children were tested individually in a quiet room at their educational institution within a single experimental session. The study was conducted in the children’s native language (Polish). After familiarising children with wearing a blindfold (eye masks for kids) and using a Swedish raised-line drawing kit (they were allowed to use one foil and draw on it whatever they wanted; they were also encouraged to control the drawings by touch – in the foil blindfolded condition), participants were asked to draw eight objects in four blocks: (1) on paper, sighted, (2) on paper, blindfolded, (3) on foil, sighted, and (4) on foil, blindfolded. Hence, the total number of produced drawings was 32. The order in which blocks were presented was counterbalanced within participants and age groups (there were 24 orders, one per participant in each age group). Within each block, participants were asked to draw eight objects: (1) a key, (2) an apple, (3) a shoe, (4) a backpack, (5) a dog, (6) a floor lamp, (7) a car, and (8) a house (sizes of these objects in real-world increase logarithmically – see 2; these objects varied in size rank), one object per page. The order of drawing was randomised within blocks. Sheets of paper/foil were horizontally arranged for the drawing task. Children drew from memory using a pencil, with no restricted time for drawing. In the blindfolded condition, children put on blindfolds before they started drawing to block the visual control. Importantly, in previous studies among preschool-aged children, blindfolds were used successfully 26,29,33,34. The experiment lasted approximately 50–60 min.

Data analyses

The aim of this study was to investigate the predictors of drawn size, operationalised as the diagonal length of drawings measured in millimetres, considering the between-subjects factor of age group (5, 7, and 9-year-olds) and three within-subjects factors: material used (paper, foil), perceptual condition (sighted, blindfolded), and size rank (1–8).

The statistical analysis was carried out in RStudio using a range of packages, each contributing to different stages of the analysis pipeline. For reading and importing the data from Excel files, we utilized the readxl package35. Data wrangling, including filtering, reshaping, and organizing the dataset, was performed using dplyr36 and tidyr37, which allowed us to prepare the dataset in a tidy format necessary for subsequent analysis.

Our preliminary analyses involved testing linear mixed-effects models, for which we employed the lme4 package38. This helped us establish an initial understanding of the fixed and random effects at play before moving on to the more complex Bayesian analysis. Additionally, glmmTMB39 was briefly tested for fitting generalized linear mixed models during exploratory analyses, particularly when dealing with potential non-Gaussian family distributions.

For the Bayesian multilevel models, we used the brms package40, which facilitated fitting complex models, including interactions between age group, material, perceptual condition, and size rank, with random intercepts for participants (ObservationID). The models were fit using eight Markov Chain Monte Carlo (MCMC) chains, running for 6,000 iterations, with 3,000 iterations allocated for warm-up, ensuring convergence. Diagnostic checks were performed using the posterior package41, which allowed for a detailed examination of the posterior samples, including trace plots and density plots to confirm the proper convergence of the MCMC chains. These diagnostic plots can be generated within the RStudio script (see the link in the Data availability section).

To explore the effects of interaction terms and marginal means, the emmeans package42 was employed. This package enabled us to conduct post-hoc pairwise comparisons and estimate marginal means, which were essential for interpreting significant interactions between the predictors.

Visualisation of the data was crucial in understanding and communicating the results. For this, we extensively used ggplot243, which was employed to generate all the figures in the manuscript, including the dot plots and violin plots that help depict the distribution of drawn sizes across different conditions.

In terms of hypothesis testing, BayesFactor44 was used to compute Bayes factors for additional Bayesian hypothesis testing, while lmerTest45 was incorporated to provide p-values for the linear mixed-effects models when necessary. Furthermore, sjPlot46 and sjmisc47 were used to generate tables and visual summaries of model results, making it easier to present the findings clearly. Lastly, we used writexl48 to export the results into Excel format for further analysis and reporting.

Results

Main analysis: size of drawings

The Bayesian multilevel model indicated substantial variability in the overall drawing size between individuals, with the random intercept for participant showing an estimated standard deviation of 79.84 (CI [67.64, 94.75]). This result underscores the importance of accounting for participant-level differences within a hierarchical model structure. Regarding fixed effects, only the significant ones will be described (for all effects for the tested model, see Supplementary Materials 1, Table S1). Descriptive statistics for drawn size are summarised in Table 1.

Table 1 Descriptive statistics of drawn size (in mm) for all experimental conditions: means and standard deviations (presented in parentheses).
Full size table

The effect of perceptual condition on drawn size showed strong evidence, with participants in the sighted condition producing larger drawings compared to those in the blindfolded condition, B = 21.12, SE = 6.68, 95% CI [7.98, 34.17] (here and hereafter in the main manuscript and supplementary information B = unstandardized posterior mean estimate and BF = BF10 from the Savage–Dickey ratio test; see Fig. 1). The Bayes Factor (BF = 2871.12) indicated very strong evidence in favor of the perceptual condition’s effect.

Fig. 1
figure 1

Drawing size as a function of perceptual condition. Violin plots, coloured coral and cyan for “Blindfolded” and “Sighted” conditions, respectively, depict the distribution of drawing sizes for each perceptual condition. Means and error bars are placed inside the corresponding violins; the red diamond represents the mean, and the error bars signify 95% confidence intervals computed using the bootstrap method. Individual observations are displayed as slightly faded points scattered within each violin plot.

Full size image

A linear effect of size rank was observed, B = 73.00, SE = 13.36, 95% CI [46.77, 99.13], indicating that as the size rank increased, participants drew larger objects (see Fig. 2). The Bayes Factor (BF = 279704692155.51) provided extremely strong evidence supporting the linear effect. Non-linear effects were also detected, with a notable negative effect for higher-order contrasts, particularly for the sixth- (BF = 5263.15) and seventh-degree (BF = 51.92) polynomials, suggesting deviations from the overall linear trend (see the findings regarding the interaction between size rank and perceptual condition presented further in this section).

Fig. 2
figure 2

Drawn size as a function of object size rank with multiple trend lines. Violin plots depict the distribution of drawing sizes for each rank level, showing the spread of data. Individual observations are displayed as points scattered within each violin, with a slight transparency for better visibility. The red diamond represents the mean for each rank, and the black error bars signify 95% confidence intervals computed using the bootstrap method. Three trend lines are added to illustrate different levels of polynomial fitting: the dashed blue line represents a linear trend, while the solid green line and solid purple line indicate sixth-degree and seventh-degree polynomial trends, respectively. These trend lines provide a visual comparison of how different polynomial fits align with the data.

Full size image

Several interaction effects were observed, two of which were related to age. First, the interaction between age group and the material used was observed, B = − 24.12, SE = 9.44, 95% CI [− 42.57, − 5.66], indicating that older children drew smaller objects when using paper compared to foil. This effect was significant only for the 9-year-olds (see Fig. 3), with a Bayes Factor (BF = 7.05) providing moderate evidence for this interaction. Moreover, an interaction effect was found between the age group and the perceptual condition, B = 22.39, SE = 9.49, 95% CI [3.92, 41.06], showing that the 9-year-olds in the sighted condition drew larger objects compared to the 7-year-olds (see Fig. 4). The Bayes Factor (BF = 393.73) indicated strong evidence supporting this interaction.

Fig. 3
figure 3

Drawn size as a function of age group and material. Violin plots illustrate the distribution of drawn sizes for each combination of age group (7 or 9 years old) and material (foil or paper), with groups separated for clarity. The black dot within each violin represents the mean, and the error bars denote 95% confidence intervals calculated using the bootstrap method. Individual observations are shown as points scattered between groups, coloured according to the material.

Full size image
Fig. 4
figure 4

Drawn size as a function of age group and perceptual condition. Violin plots display the distribution of drawn sizes for each age group (7 or 9 years old) under two perceptual conditions (blindfolded and sighted), with groups separated for clarity. Means and error bars are positioned above the corresponding violins, where the black dot represents the mean and the error bars indicate 95% confidence intervals calculated via the bootstrap method. Individual observations are presented as points scattered between groups, coloured according to the perceptual condition.

Full size image

Another significant interaction relates to the interaction between the perceptual condition and size rank factors. Specifically, within the visual perceptual condition, participants exhibited a larger increase in drawn size as size rank increased (see Fig. 5), with evidence observed for both linear, B = 46.76, SE = 18.80, 95% CI [9.89, 83.42], BF = 4168.47 and quadratic, B = 46.33, SE = 19.02, 95%CI [9.35, 83.36], BF = 218.23 trends. To gain a deeper understanding of the relationship between size rank and drawn size, we extended our analysis to explore whether this relationship was better described as linear or quadratic under different perceptual conditions. Specifically, we aimed to determine if the sighted and blindfolded perceptual conditions differed in the type of relationship (linear or quadratic) best describing the data (see Supplementary Materials 1, Table S2 and Table S3). The Leave-One-Out (LOO) cross-validation analysis indicated that the relationship between size rank and drawn size was best explained by a quadratic relation when drawings were produced under visual control, while a linear relationship provided a slightly better fit than a quadratic one when participants drew without visual control.

Fig. 5
figure 5

Drawn size as a function of object size rank and perceptual condition. Violin plots illustrate the distribution of drawn sizes for each size rank, separated by perceptual condition (coloured coral for “blindfolded” and cyan for “sighted”). Individual observations are displayed as points scattered within each group, coloured according to the perceptual condition. The black dot within each violin represents the mean, and the black error bars denote 95% confidence intervals calculated using the bootstrap method. Two types of trend lines are included for each perceptual condition: the dashed line represents a linear trend, while the solid line represents a quadratic trend. Both trend lines are shown with shaded confidence intervals to illustrate the precision of the fit.

Full size image

Additional analysis: recognisability of drawings

The size of the drawings may have been affected by the occurrence of positioning errors in the drawings related to the lack of accuracy in rendering the object’s topology5,29. Since the quality of drawings produced by children under different perceptual conditions increases with age29, and drawings characterised by higher quality may contain fewer positioning errors5, it was worth considering drawing quality as a confounding variable in the analysis of the canonical size effect among children. Therefore, we sought to investigate whether the recognizability of children’s drawings (assessed by judges within an additional study – see Supplementary Material 2 Methods and Table S4) mediated the relationship between these factors and drawing size, conducting a multilevel Bayesian mediation analysis (see Supplementary Material 2 Results and Figure S1). The findings indicate that recognisability substantially mediates the relationship between the predictors (age, perceptual condition, size rank, and material) and the size of the drawings. Although recognisability was associated with drawn size, the retention of variance being explained for the key predictors in the direct path indicates only partial mediation. When comparing the R² values, the total model, which includes both direct and indirect paths, explained 66% of the variance in drawing size, R2 CI [0.64; 0.67]. The direct path accounted for 35% of the variance, R2 CI [0.34; 0.36], while the indirect path explained 31% of the variance, R2 CI [0.29; 0.33].

Discussion

To test the canonical size phenomenon among children, we investigated three age groups (5-, 7-, and 9-year-olds) in a drawing-from-memory task. Drawings of objects whose size increases in the real world were produced on sheets of regular white paper or special foil for embossed drawings. When creating drawings, children were sighted or blindfolded, depending on the condition.

In line with hypothesis 1, the size rank (which refers to the real-world size of a given object in a logarithmic way2) showed strong evidence of influencing the drawn size. Irrespective of the perceptual condition under which the drawings were produced, the material used for drawings, and the children’s age, the relation between size rank and drawn size was positive. At the same time, although the overall character of this relation was linear, the pattern of results turned out to be more nuanced when investigated distinctively within the sighted and blindfolded perceptual conditions (as follow-up analyses after revealing significant interaction between perceptual condition and size rank). In the blindfolded condition, the linear function best explained the relation between size rank and drawn size – which replicated the findings previously shown among blindfolded adults5 and blind participants15. In the sighted condition, a quadratic function provided stronger evidence for describing the relationship between size rank and drawn size compared to a linear function. Importantly, this quadratic pattern of results was caused by different slopes of the function because less incremental drawn size with increasing size ranks revealed for lower than higher size ranks. Still, the drawn size was larger for higher than lower size ranks of objects. Thus, we may conclude that in both perceptual conditions, a canonical size effect was found in children aged 5 to 9 when drawing from memory. This means that children, like adults, are sensitive to systematically varying differences in the size of real-world objects and that information about the real-world size of an object is an essential aspect of its mental representation2. Notably, in the original study by Konkle and Oliva2, the differences in the given size ranks were intentionally scaled logarithmically. Consequently, it is reasonable to speculate about the relationship patterns between inner size representation and actual size among children. A linear relationship between size rank and drawn size may indicate a logarithmic scaling of the size representation in memory – which was the case in the blindfolded condition. In contrast, a quadratic function – yielded in the visual perceptual condition – may indicate a linear scaling.

Furthermore, similar to adults (see, e.g.5,15), children as young as 5 can draw from memory even without visual control, most likely relying either on a procedural or an abstract declarative memory of shape and size and specific (habitual) patterns of movements when drawing the objects. This means that these automatised movements include information about the range of movement that the hand must make to reflect the real-world size information represented in long-term memory. We can speculate that a relatively similar pattern of results with respect to the canonical size effect tested in the visual and blindfolded perceptual conditions speaks for the spatial nature of the canonical size phenomenon (which may be operationalised by the drawing task, among others) and, more generally, inner size representation, as suggested in previous studies conducted in the haptic domain among adults4,5,6 and eventually evidenced in the research among blind adults15. The current study allows us to generalise these conclusions beyond adults, as it seems that by age 5 to 9, real-world object representation (or at least information about an object’s size) is represented spatially rather than visually. Besides, there was no substantial evidence indicating that the canonical size effect was moderated by participants’ age. Therefore, it seems that between the ages of 5 and 9, there are no noticeable developmental differences in representing information about the size of objects that are increasing real-world size.

Our study revealed some other interesting effects besides showing the canonical size phenomenon in children. As recently shown among adults4,5, children produced larger drawings within the sighted perceptual condition than in the blindfolded perceptual condition. Future studies need to explore whether this effect was caused by the children’s attempts to control the drawing process and minimise the effort of working memory when drawing without visual control—as suggested in a study among adults5. Unfortunately, the process of producing drawings itself was not measured in the current experiment. Future studies may attempt to fill this gap, for example, by using neuroimaging techniques.

Two interactions – between age and material used for drawing, as well as between age and perceptual conditions – shed some light on the developmental differences when producing drawings under circumstances that are more and less familiar to children. The findings showed that 9-year-olds produced larger drawings than 7-year-olds in the sighted perceptual condition. This may be due to the more details produced in their drawings of a given object by older children (i.e., those who are more experienced and trained in drawing on paper) than younger ones49. Moreover, children aged 9 years (but not two other age groups of younger children) drew smaller objects when using paper compared to foil. Probably, 9-year-olds had more drawing practice and sensorimotor training than younger tested children and consequently could produce tinier drawings of objects when using familiar material (like a standard paper sheet) compared to unfamiliar material (like a special foil for producing embossed drawings).

Finally, the recognisability of drawings produced by children showed strong evidence of being correlated with drawn size, and recognisability partially mediated the relationship between the key predictors (such as size rank) and drawn size. Therefore, in future research on the canonical size phenomenon, the recognisability of drawings should be considered and included in the analyses. Ideally, this variable should be supplemented with other potential confounding variables related to the produced drawings, such as the number of details or perspective of presenting a given object.

The current study has certain limitations. First, only the drawing from memory task was used in the study. In the original research on the canonical size effect2, the adult participants were also tested in visual perception (tasks measuring perceptual preference were used) and imagery (participants were instructed to imagine a given object when looking at a blank screen and then mark a tight bounding box around the mental image of that object) experiments. Although not all of the tasks used in the original research seem to be fitting for children (e.g., the imagery task may be challenging to understand for preschoolers), future studies on the canonical size phenomenon among children could be extended to include tasks on the preference of images of objects varied in size in the real world presented in different sizes in a constant-size frame (cf.2). Second, the current study used a mixed paradigm, with a between-subjects factor of age and within-subjects factors of perceptual condition and drawing material. The disadvantages of a within-subjects design, especially regarding younger children, are the relatively long duration of the experiment and the conservatism (i.e., relative inflexibility caused by children’s repeated drawings)50 in the drawings produced in varied perceptual and material conditions. The third group of limitations refers to the changes in the drawing from memory task’s method when compared to the original one2. We adopted the procedure of Konkle and Oliva2 without conducting a preliminary study which aimed to assess the size rank of objects among children. It limits our conclusions because we cannot be sure whether the size rank determined in a study of US adults would be replicated if a similar preliminary study was conducted among Polish children. Moreover, we changed the procedure originally used by Konkle and Oliva2 by using only one object to draw per size rank category (originally, there were two objects per category) and describing some objects less precisely (e.g., dog instead of German Shepherd, shoe instead of running shoe). Similar modifications to the original procedure were previously applied in studies on the canonical size effect among adults5,15, which replicated the findings of the original research2. Nevertheless, it is undoubtedly a significant limitation of the current research. Besides, Konkle and Oliva 2 showed that canonical size is revealed independently of the size of the sheet of paper on which objects are drawn. As we only used one size of drawing material, we cannot be sure to what extent the children were adapting the size of the objects they were drawing to the frame of the paper sheet/foil rather than presenting the habitual size (the size they usually draw for the given objects20) in their drawings.

This study also provides perspectives for future research. The study was conducted among sighted children. In future studies, it would be worthwhile to additionally examine blind children in a similar age range – to assess whether the presence of the canonical size phenomenon in children is moderated by their everyday experience of seeing objects from different distances and experiencing changes in visual angular sizes. While the canonical size effect has already been observed in congenitally blind adults15, it is still unclear how early in development this phenomenon occurs among individuals lacking visual experience.

To conclude, this study has shown for the first time that the phenomenon of canonical size occurs in children from the age of 5, both when investigated in the visual modality and when drawings were produced without visual control, irrespective of the children’s age and material used for drawing. Hence, our study suggests the robustness of this phenomenon among children because it was observed when producing drawings with varied perceptual feedback provided – visual, haptic, or even proprioceptive only (see5).

Data Availability

The data sets generated and analysed during the current study, alongside the dataset codebook and the R script used to perform all analyses, and reproductions of drawings produced by the participants, are available in the Open Science Framework repository (https://osf.io/kcnga/?view_only=bdfd930e20ea48e286265ddb09765e07).