Children's picture interpretation: Appearance or intention?

Experiment 1

In Experiment 1 we explored whether children think appearance or intention is more important for interpreting pictures. We used a between subjects design; children took part in the photograph or the line drawing task and in the color change or the black and white condition. In both conditions we changed the appearance of a series of pictures to create conflict between what the picture creator intended to depict and what the picture resembled. In the color change condition the color of the referent was changed (e.g., if the experimenter intended to draw a blue duck, the final picture showed a pink duck) and in the black and white condition the picture appeared in grayscale, rather than color. The modality manipulation also allowed us to examine whether children’s cue use differs for photographs and drawings.

One-hundred and 51 typically developing children between the ages of 3 and 6 participated in the photograph ( N = 76) and line drawing tasks ( N = 75). Children were split into two age groups: 3- and 4-year-olds ( M age = 46 months; Range = 37–59 months) and 5- and 6-year-olds ( M age = 71 months; Range = 60–82 months), and two conditions, yielding four experimental conditions (see Table 1 ). Children were recruited from six primary schools, three nurseries, one holiday play scheme, and the database of the Centre for Research in Human Development and Learning (CRHDL) at Lancaster University. Families were predominantly White and middle class.

Line drawing task

Eight coloring crayons and plain A4 paper were used. To permit direct comparisons across tasks the same objects, arranged in the same three-array configurations as in the photograph task were used here. Pictorial stimuli comprised 10 color or grayscale line drawings (8 in. × 4 in. and presented landscape on A4 paper) of these objects.

Design. A 2 × 2 × 2 × 3 mixed design was used. Condition (“color change” and “black and white”), modality (“photograph” and “drawing”), and age group (“3- and 4-year-olds” and “5- and 6-year-olds”) acted as the between-subjects factors. The within-subject factor was question type (“verbal,” “behavioral,” and “memory control”). Intentional responses per question type were summed across trial to form three composite scores (see Coding section).

Line drawing task

Children participated in four test trials. The procedure followed that of the photograph task with some minor instructional changes. Children were told the experimenter was going to draw one of the objects (“I am going to draw a picture of the blue duck”). In the color change condition, when choosing a crayon the experimenter looked at the selection (which included both correct and incorrectly colored crayons) and said, “I’m going to use this one” before picking up the “wrong color” crayon. For instance, if the experimenter chose to draw the blue duck, she picked up the pink crayon. In the black and white condition the crayon chosen was always black. Highlighting the creator’s intention to choose a particular marker (in this case, a specific colored crayon) is a method that has been successfully used in previous work to assess children’s understanding of the role intention plays in representation ( Myers & Liben, 2008 ).

Coding

In the color change condition, responses were coded as either intentional, appearance, or “other.” Responses were coded as appearance-based if the child’s label for, or physical choice of object, matched the color of the object in the photograph or line drawing. Intentional codes were assigned if the child’s physical object choice or verbal label matched the color of the object initially photographed. The “other” code was reserved for responses that did not conform to either of the above response types, for instance, choosing the distractor object.

In the black and white condition, responses were coded as intentional, nonintentional, or “other.” Intentional codes were assigned if the child’s object choice or label matched the object that was initially photographed. Nonintentional codes were assigned if children’s object choice or label matched the object that was not photographed (e.g., choosing the pink duck when the experimenter had intended to photograph the blue duck). Appearance codes were not utilized here as selecting the nonintended object (e.g., pink duck) did not match the grayscale appearance of the picture. “Other” responses included the distractor object, because this was the only additional response ever provided. The same coding scheme was used for the photograph and line drawing tasks.

Results and Discussion

To provide an initial view of any patterns in the data, the percentage of question responses falling into each of the three coding categories was calculated. The majority of responses given by participants fell into the “appearance/nonintentional” and “intentional” categories. Children’s responses were coded as “Other,” indicating that they chose the distractor object from the array, on a total of 6.2% of trials. Due to the infrequency of these responses they were removed from subsequent analyses, which focused on comparing appearance/nonintentional and intentional question responses.

Children each had 12 data points, having answered three questions per trial across four trials. Although different stimuli were used on each trial, and trial order was counterbalanced, it was important to check that children’s question responses did not differ as a function of the stimuli used, or the order in which they were presented. McNemar tests were conducted to identify possible stimulus effects, as no such effects were identified the data were collapsed across trials. A repeated measures ANOVA was then used to check for order effects; none were identified. The dichotomous nature of the “appearance/nonintentional” and “intentional” response categories necessitated that only one response type act as the dependent variable. Thus, intentional question responses were chosen and summed across all four trials to provide three composite scores, one per question type: verbal, behavioral, and memory control. The final DV was number of intentional responses out of four trials, thus scores ranged from 0–4. A score of 0 indicated that no intentional responses were given to that question, whereas a score of 4 indicated that intentional responses were given to all questions of that type.

To check whether children had remembered the experimenter’s original intention responses to the memory control question were analyzed first. Those children who responded correctly to the memory control question on three or four trials were placed in the “passed memory control” group, and those who responded correctly on zero, one, or two trials were placed in the “failed memory control” group. A total of 90 children remembered, while 61 forgot (see Table 2 for relevant age and condition groupings). Due to the large number of children who had apparently forgotten the experimenter’s stated intention, responses from children who passed and failed the memory control were analyzed separately using individual 2 (Modality: photograph, line drawing) × 2 (Condition: color change, black and white) × 2 (Age group: 3- and 4-year-olds, 5- and 6-year-olds) × 2 (Question Type: verbal, behavioral) repeated measures ANOVAs.

A significant main effect of question type, F (1, 82) = 9.62, p = .003, η p 2 = .11, revealed that children gave more intentional responses to the behavioral ( M = 2.30, SE = .13) than the verbal question ( M = 1.89, SE = .14). A significant main effect of modality, F (1, 82) = 5.77, p = .019, η p 2 = .07, also indicated that children in the line drawing task ( M = 2.40, SE = .16) gave more intentional responses than children in the photograph task ( M = 1.80, SE = .18). Significant main effects of condition, F (1, 82) = 87.47, p < .001, η p 2 = .52, and age group, F (1, 82) = 4.28, p = .042, η p 2 = .05, were qualified by a Condition × Age Group interaction, F (1, 82) = 9.09, p = .003, η p 2 = .10 (see Figure 2 ). In order to establish the nature of this interaction, additional repeated measures ANOVAs were conducted on data from the two age groups and two conditions separately. These revealed that in the black and white condition, 3- and 4-year-old children ( M = 3.16, SE = .28), F (1, 37) = 12.88, p = .001, η p 2 = .26, and 5- and 6-year-old children ( M = 3.31, SE = .20), F (1, 49) = 114.17, p < .001, η p 2 = .26, gave significantly more intentional responses than children in the color change condition (3- and 4-year-olds: M = 1.62, SE = .32; 5- and 6-year-olds: M = .37, SE = .19). However, in the color change condition, 3- and 4-year-old children ( M = 1.58, SE = .28) gave significantly more intentional responses than the 5- and 6-year-olds ( M = .36, SE = .22), F (1, 47) = 11.79, p = .001 η p 2 = .23.

Mean number of Intentional responses given in the color change and black and white conditions by children who passed the memory control.

Chance analyses (chance value = 2), conducted using one sample t tests, further revealed that in the black and white condition, 3- and 4-year-old children gave significantly more intentional responses to the verbal ( M = 2.95, SE = .33), t (21) = 2.87, p = .009, and behavioral questions ( M = 3.36, SE = .23), t (21) = 5.85, p < .001, than would be expected by chance, as did the 5- and 6-year-olds: verbal ( M = 3.13, SE = .32), t (23) = 3.51, p = .002; behavioral ( M = 3.50, SE = .24), t (23) = 6.23, p < .001. In the color change condition, 3- and 4-year-old children performed at chance on both the verbal ( M = 1.35, SE = .37), t (16) = −1.73, p = .10, and behavioral questions ( M = 1.88, SE = .43), t (16) = −.28, p = .79, while 5- and 6-year-old children gave significantly fewer intentional responses to the verbal ( M = .22, SE = .12), t (26) = −14.42, p < .001, and behavioral questions ( M = .25, SE = .20), t (26) = −7.59, p < .001, than would be expected by chance.

Failed memory control

A main effect of condition, F (1, 53) = 56.14, p < .001, η p 2 = .51, revealed that overall children in the black and white condition ( M = 3.12, SE = .16), gave more intentional responses than children in the color change condition ( M = 1.36, SE = .18). One-sample t tests confirmed that children in the black and white condition gave more intentional responses than would be expected by chance, verbal: t (30) = 4.60, p < .001; behavioral: t (30) = 6.77, p < .001, whereas children in the color change condition gave fewer intentional responses than would be expected by chance, verbal: t (29) = −4.57, p < .001; behavioral: t (29) = −3.80, p = .001.

A main effect of age group, F (1, 53) = 17.34, p < .001, η p 2 = .25, was qualified by a Modality × Age Group interaction, F (1, 53) = 6.98, p = .011, η p 2 = .12 (see Figure 3 ). In order to establish the nature of this interaction, additional analyses were conducted on the two tasks and two age groups separately, which revealed that in the photograph task, 5- and 6-year-old children ( M = 3.27, SE = .38) gave significantly more intentional responses than the 3- and 4-year-olds ( M = 1.26, SE = .27), F (1, 32) = 21.74, p < .001, η p 2 = .41. No such effect was identified in the line drawing task.

Mean number of Intentional responses given in the photograph and line drawing tasks by children who failed the memory control.

In summary, a different pattern of results emerged for children who passed and failed the memory control. Intriguingly, the performance of children in the latter group does not necessarily support the notion that they had forgotten the experimenter’s intention. If these children genuinely did not remember which of the three objects was the intended referent we would have expected them to perform at chance, or to rely consistently on appearance cues when answering the verbal and behavioral questions. Instead, they showed the same bias toward intentional responding in the black and white condition as children who passed the memory control, while the 5- and 6-year-olds in the photograph task also relied on intentional cues to identify the picture’s referent. An alternative explanation is that these children were simply not convinced that the experimenter “meant” to draw or take a picture of the referent she had identified, and thus failed to report her stated intention when asked the memory control question due to its intentional phrasing; “What did the experimenter mean to draw a picture of” rather than “What did the experimenter say she would draw a picture of?” In other words, while children were unwilling to explicitly endorse the idea that the experimenter had truly intended to draw or photograph one object while producing a picture that contradicted her intention, they were willing to use her statement of intent, in the black and white condition, to identify the picture’s referent.

Returning to those children who did remember the experimenter’s intention, the current findings demonstrate that the referential ambiguity of a picture is of fundamental importance in determining the use of appearance and intentional cues. When an image could represent multiple referents and therefore appearance cues were unavailable, as in the black and white condition, the picture’s identity is ascertained using intentional cues. By contrast, when an image is intended to represent one object but strongly resembles another, as in the color change condition, appearance cues dominate older children’s picture interpretation. Interestingly, 3- and 4-year-old children performed at chance in this condition, indicating that they may be less dependent on the ambiguity of a picture than their older counterparts. As their performance did not exceed chance, intention did not override the picture’s appearance, instead it appears that the younger children were simply less willing to disregard the experimenter’s intention. This may be because they have only recently begun using intentional cues to interpret pictures ( Bloom & Markson, 1998 ; Gelman & Ebeling, 1998 ; Preissler & Bloom, 2008 ) and are therefore more attuned to them than older children, who have begun to tailor their cue use to the context in which they encounter pictures. Nonetheless, overall these findings pinpoint picture ambiguity as an important mediating factor in children’s developing picture interpretation.

Beyond the effects of condition and age, children who passed the memory control also gave more intentional responses to the behavioral than the verbal question. This supports the earlier hypothesis that the two questions tap different aspects of dual representation, which refers to the notion that pictures can be thought of as both objects in their own right and as representations of other entities. Corroborating our expectations, asking participants to retrieve the picture’s object referent (“Can you pass me this?”) served to remind them of the creator’s intention to depict a specific referent (“I’m going to take/draw a picture of the pink duck”), thereby highlighting the picture’s identity as a symbol that represents something in the real world. By contrast, when asked the verbal question (“What is this a picture of?”), participants could simply look at the picture itself, as an object in its own right, and name it accordingly. Finally, and as predicted, children in the line drawing task gave more intentional responses than children in the photograph task. One explanation for this is that the iconic nature of the photographs focused children’s attention on the picture-world relationship, resulting in an increase in their use of appearance cues and a corresponding decrease in intentional responding. Certainly, prior studies have found that when asked to categorize and evaluate photographs 7- to 13-year-old children largely focus on the visual properties of photographs, mainly content, to the detriment of any in-depth consideration of the photographer’s role in choosing the content

Experiment 2

The aim of Experiment 2 was to clarify the underlying motives for the predominance of appearance responses given to the verbal and behavioral questions by children in the color change conditions of Experiment 1. We hypothesized that these responses may have been prompted by the artist’s failure to comment on the picture’s changing appearance, thereby leading children to believe that her earlier intention was no longer relevant. Previous research has shown that young children are aware that what people see directly affects their knowledge of objects or events ( O’Neill, Astington, & Flavell, 1992 ; Pratt & Bryant, 1990 ; Pillow, 1989 , 1993 ; Robinson & Whitcombe, 2003 ; Wimmer, Hogrefre, & Perner, 1988 ) and can use this information to distinguish between knowledgeable and ignorant observers ( Einav & Robinson, 2011 ; Koenig & Harris, 2007 ; Robinson, Butterfill, & Nurmsoo, 2011 ). In order to explore whether the knowledge of the experimenter influenced children’s picture choices in Experiment 1 we added a second experimenter. Experimenter 2 knew what the final picture looked like, but did not know anything about Experimenter 1’s (the artist) intentions. Conversely, while Experimenter 1 knew what she intended the picture to represent she never saw the final image. Half the children were asked the test questions by Experimenter 1, and the other half were asked by Experimenter 2. We expected children to accept that the knowledge of the two experimenters did not overlap, as previous work has shown that much younger children do not expect knowledge acquired by one person to be known by another ( Moll, Richter, Carpenter, & Tomasello, 2008 ; Tomasello & Haberl, 2003 ). Only the color change condition was used in Experiment 2, as it was in this condition that children were guided by appearance cues.

It was predicted that, if children considered the artist’s knowledge when interpreting the picture, then when Experimenter 1 (the artist) asked the questions, children should give predominantly intentional responses as Experimenter 1 only knew what she intended to depict. When Experimenter 2 asked the questions, children were expected to give appearance responses because Experimenter 2 only knew what the picture looked like. Contrastingly, if the children were staunch realists they were expected to give appearance responses regardless of which experimenter asked the test questions, because the picture’s appearance did not change across conditions. No age group differences were expected.

Eighty typically developing 3- to 6-year-old children participated. They were split into two age groups: 3- and 4-year-olds ( M age = 50 months; Range = 40–59 months) and 5- and 6-year-olds ( M age = 70 months; Range = 60–82 months). See Table 3 for condition and age groupings. Children were recruited from four primary schools in North Yorkshire, and the database of the Centre for Research in Human Development and Learning (CRHDL) at Lancaster University. Families were predominantly White and middle class.

Apparatus and stimuli

The materials were identical to the color change condition of Experiment 1; however, due to the nature of the procedure, only two of the original four trials were included. Pilot testing revealed that children would not believe that a second experimenter would interrupt with four “unexpected” phone calls during the short testing session. As no stimulus effects were found in Experiment 1, the duck and spoon trials were randomly chosen.