Article Text

Download PDFPDF

What features of images affect parents’ appraisal of safety messages? Examining images from the A Million Messages programme in Canada
  1. Barbara A Morrongiello1,
  2. Melissa Bell1,
  3. Michael Butac2,
  4. Alexa Kane1
  1. 1Psychology Department, University of Guelph, Guelph, Ontario, Canada
  2. 2Mississauga, Ontario, Canada
  1. Correspondence to Dr Barbara A Morrongiello, Psychology Department, University of Guelph, Guelph, ON, Canada N1G 2W1; bmorrong{at}


Background Enhancing caregivers’ awareness of children's injury risks and increasing knowledge about strategies for injury prevention often involve presenting parents with written materials and accompanying images.

Objectives To assess parents’ appraisals of different variations of images and identify those features that enhance their attention to safety messages.

Methods Eight images showing risk situations were taken from the A Million Messages safety education parent-directed programme in Canada and modified to create a corresponding image that clearly showed negative consequences for the child, and facial expressions of fear and/or upset. Mothers with young children were presented with the eight pairs of images (negative consequence vs risk situation) and asked to select the best accompaniment to a safety message and to provide an explanation for their choice. Each image was then also rated for fit to the safety message, communication of danger, emotional arousal and attention elicitation.

Results The images depicting negative consequences were chosen for most comparisons (78%) and higher scores were assigned to these images for all four features rated by parents (danger communicated, emotions evoked, attention elicitation and fit to the safety message). Moreover, ratings of danger, emotions and attention predicted ‘fit to safety message’ scores.

Conclusions Depicting negative consequences and showing negative emotions is important to maximise the effectiveness of images in communicating danger and evoking attention and concern when targeting parents with child-safety messaging.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Unintentional injuries are the leading cause of death and disability during childhood.1 ,2 Owing to their lack of understanding of risk and their limited physical/motor and psychological development, young children depend on caregivers to protect them from hazards.3 Many intervention programmes have been developed to increase caregivers’ awareness of risks and strategies to prevent childhood injuries. Often these communicate their messages using written text with accompanying images. Past research highlights effective strategies for communicating written safety messages.4 However, there has been surprisingly little research evaluating parents’ perceptions and interpretations of images that accompany safety messages. This study addresses this gap in the literature.

Research has shown that images used in warnings can attract attention, convey information and promote memory of this information.5 ,6 Taylor and Bower reported that the impact of a product warning is improved when an accompanying image is vivid (ie, easily imagined, visualised, interesting, thought-provoking).7 Moreover, increased ability to visualise the consequences of a written message enhances one's perception that the outcome is more likely. Additionally, images and symbols enhance text comprehension, which is especially important for conveying safety information to a diverse audience. People with low literacy levels or non-native speakers, for example, often have difficulty in understanding safety and warning messages when detailed text is presented. However, supplementing text with images can enhance understanding of health messages.8 Thus, images that are well suited to written messages can have the beneficial effect of evoking attention to, and memory of, the message for a more diverse population.

Building on these findings, this study assessed parents’ appraisals of different variations of images and aimed to identify features that enhance the impact of safety messages on parents. To develop contrasting images, we first identified a popular community-based programme in which images play an integral role in communicating safety messaging. A Million Messages (AMM) is an innovative and popular programme developed in Alberta, Canada that is implemented through public health departments and partly uses written media to provide families with child injury prevention messages.9 Evaluation of AMM indicates that parents report that the safety messages provided new information (66%) and had changed their behaviour to some extent (46%).10 Interestingly, the images depict risk situations but there is variation across images in the affect communicated (eg, some show the child communicating a positive emotion in a risk situation) and the extent to which negative consequences of injury are depicted. This kind of variation in imaging is common in safety education materials for parents, and shows the need for research to identify those features that maximise impact on parents’ appraisals of injury risk and motivations to implement safety precautions.

In this study, we selected AMM images that showed important injury risk situations for young children and then created versions of these that depicted negative consequences of injury and showed the child upset, with the aim being to determine if these features significantly improve parents’ appraisal of injury risk relative to images that showed only the risk situation and/or communicated positive emotions. We focused on these features because research indicates that emphasising potential consequences can be effective in convincing individuals there is a true risk of injury or harm.11–13 Research has shown too that depicting consequences that violate a person's expectation of safety can be particularly important if the warning message applies to a highly familiar context in which the person has not previously experienced injury.14 Finally, showing a child upset can evoke parent attention and emotional reactions,15 which might also increase the impact of safety messages on parents. Kelly, Sein and McCarthy,16 for example, concluded that safety messaging must be emotionally engaging in order to change parents’ perceptions and have an enduring effect on their behaviour.


Study design

Based on findings from research on the impact of negative images of injury on parents’ appraisals of child-injury risk,17 we expected to obtain medium to large effect sizes (ie, ηp² of 0.38 to 0.59) when conducting analyses to compare means across different image conditions. With α level set at 0.05 and power set at 0.80, this resulted in an estimated sample size of between 26 and 60, based on Cohen.18


The sample comprised 60 mothers (Mage=35.97 years; SD=4.71 years) of young children (M=4.41 years, SD=0.80 years), with participants randomly selected from a database of volunteer families from the local community who were interested in child development research. Most mothers were Caucasian (93%), had a college or university degree (55%) and had an annual family income above $C80 000 (55%).


Eight safety messages were taken from the AMM program. Each message was presented with a pair of images, one depicting a risk situation (original) and the other a modified version that depicted negative consequences (injury, crying child). Each pair of images was presented on a white poster board with Velcro backings to make it easy to switch images. A sample of four contrasting images is shown in figure 1. It should be noted that the images shown are black and white but the 16 images used in the study (eight showing the risk situation and eight showing the negative consequence) were coloured just as they are in the AMM programme.

Figure 1

Sample of four paired images, with each pair including the original and the modified version that emphasized negative consequences. In the study, these were presented to parents in colour.

A script for the parent interview and response recording sheet were developed to ensure that the study was conducted in a similar manner for all the participants. Additionally, randomisation sequences for image order, the four rating scales and placement of the original and the modified image (left vs right) were designated for each participant. Ratings of fit to the message, attention elicitation, communication of danger and emotional arousal were based on a six-point Likert scale (0=not at all, 1=a little, 2=somewhat, 3=fairly, 4=very, 5=extremely). Each interview was audio recorded and later, explanations for the reasons participants chose a particular image were coded.


After parents provided written consent they were shown a written safety message together with the two images. Parents selected the image that they thought best communicated the safety message and their explanations of ‘why’ were audio recorded. Parents then used the Likert scales to rate each image based on fit to the accompanying safety message, communication of danger and threat of injury, emotional arousal and effectiveness in capturing one's attention. This process was repeated for all eight pairings.

Audio coding

The audio coding scheme was developed based on examining responses by parents. Explanations for why the parent selected the image chosen were grouped into four common types and included those (1) involving emotion (eg, child looks very afraid), (2) reminding of hazards (eg, shows how dangerous open toilets are), (3) reminding of consequences (eg, shows how badly a child could get hurt), or (4) showing there were problems with the non-selected image (eg, doesn't make sense that the child is happy).

A second coder independently coded 25% of the transcripts (randomly selected) and agreement of >90% was achieved for each coding category. The data from the primary coder were analysed.

Analytical approach

Descriptive and parametric statistics were applied to characterise the data and compare the pattern of results across the two types of images (negative consequence vs risk situation). To determine if there was a preference for images showing a negative consequence or those showing a risk situation, we computed the percentage of the eight image pairs for which the image showing a negative consequence was chosen and compared this overall score with chance or 50% using a one-sample t test. To determine the most common reasons for choosing the negative consequence image, an analysis of variance (ANOVA) test was conducted comparing the proportion of each type of explanation (emotional arousal, communicates injury risk, emphasises negative consequences of injury, problem with other graphic). A significant variation in proportion scores was found and, therefore, follow-up paired comparison t tests were conducted to identify significant differences between the types of explanations given. In conducting paired contrasts using t tests, a Bonferroni adjustment for family-wise error rate was applied; the results reported are based on this adjustment.

Similarly, to identify significant differences in parents’ ratings of image characteristics between images showing a negative consequence or a risk situation, ANOVAs were conducted comparing these mean ratings for negative consequence versus risk situation images, for each characteristic separately (extent of fit of image to safety message, extent of danger communicated by the message, extent of emotional reaction to the image, extent of attention elicited by the image). While the above ANOVAs were used to determine the extent to which the ratings of the two different types of images differed, linear regressions were conducted to determine which of these four characteristic ratings predicted mothers’ actual image preferences and to what extent these characteristics did so (ie, percentage of variance accounted for).

Several preliminary data checking procedures were applied before analyses were conducted. Specifically, variables were examined for violations of normality, though none of the variables needed to be transformed. Similarly, we assessed for outliers (ie, two SD above and below the mean), but none were identified. Before reporting within-participant ANOVAs we assessed for violations of sphericity to determine if adjustment to the degrees of freedom was warranted; no such adjustments were needed. Effect sizes are reported as partial η2 because this has the advantage over η2 that the magnitude is not affected by the number of effects in the ANOVA.


Which image condition did parents prefer and why?

For each participant we computed the mean number of times that the modified (negative consequence) image was chosen across all eight safety messages and converted this to a percentage score. The negative consequence images were preferred (M=78.33%, SD=15) more so than the risk situation images. A one-sample t test confirmed that this percentage score exceeded chance level (50%), t(59)=11.25, p=0.001, Cohen's d=1.4.

A one-way ANOVA, with type of explanation as the within-participant factor, was conducted to compare explanations given for selecting the negative consequence images; the proportions of explanations of each type were analysed. As shown in table 1, there was significant variation in explanations given, F(3, 174)=5.49, p=0.001, ηp²=0.09. Follow-up paired-comparison tests showed that the proportions of explanations involving emotional arousal and communicating injury consequences were each significantly higher than the proportion of explanations related to problems with the non-selected image (p=0.048 and 0.014, respectively). Further, the proportion of explanations relating to communicating negative consequences of injury also exceeded the proportion of explanations related to communicating injury risk (p=0.024).

Table 1

Results explaining why the negative consequence image was selected

Did parents’ ratings of the four characteristics of the images vary with image condition?

For each of the four characteristics (best fit of image to safety message, danger communicated by the message, emotional reaction to the images, attention eliciting properties of the image), a separate mean score was calculated by averaging across the eight ratings parents gave for the negative consequence images and for the risk situation images, separately. A one-way ANOVA with image condition as a within-participant effect was then conducted on the data from each characteristic separately. As can be seen in table 2, for all four characteristics, the negative consequences images received more favourable ratings than the risk situations images.

Table 2

Mean (SD) ratings for negative consequence and risk situation images

Fit to the safety message ratings for the negative consequences versus risk situation conditions proved to be highly significant, F(1, 59)=300.10, p=0.001, ηp²=0.84. Ratings of sense of danger for the child for the negative consequences condition also significantly exceeded those ratings given for the risk situation condition, F(1, 59)=646.78, p=0.001, ηp²=0.92. Similar results were obtained for emotional arousal (F(1, 59)=656.79, p=0.001, ηp²=0.92) and effectiveness in capturing one's attention, F(1, 59)=381.36, p=0.001, ηp²=0.87.

What image characteristics predict parents’ ratings of ‘best fit’ of the negative consequence image to the safety message?

Linear regressions were conducted to assess for direct relations between ratings of image characteristics (communicating danger, evoking emotions, eliciting attention) and parents’ preference for the negative consequence images, and to identify those characteristics that predicted such a preference. Results of regression analyses showed that all three ratings contributed to predict parents’ preferences for the negative consequences images. Specifically, parents’ ratings of danger communicated by the modified images explained a significant proportion of variance (adjusted R²=0.73) in ratings of ‘best fit’ for modified images, F(1, 58)=157.74, p<0.001; t=12.56, p<0.001; β=0.96, SE=0.08, b=0.86. Ratings of emotion also explained a significant proportion of variance (adjusted R²=0.53) for ratings of ‘best fit’, F(1, 58)=68.25, p<0.001; t=8.26, p<0.001; β=0.67, SE=0.08, b=0.74. Similarly, ratings of attention explained a significant proportion of variance (adjusted R²=0.66) for ratings of ‘best fit’, F(1, 58)=117.13, p<0.001; t=10.82, p<0.001; β=0.91, SE=0.08, b=0.82. Hence, all three characteristics of the images showing negative consequences of injury contributed to parents judging these as a better fit to the safety message.


Although images are part of most messaging campaigns that provide parents with safety information, surprisingly few studies have sought to identify features of images that enhance the impact of safety messages. In AMM, for example, the standardised visual images are a fundamental component, which cannot be modified for this programme.9 Yet the images have varied features, with most portraying risk situations, an approach that is common to most print messages aimed at promoting safety practices among parents. Our study examined whether portraying injury consequences and displaying appropriate negative emotions on the child's face would affect parents’ perceptions of the safety messages over and above effects achieved by depicting injury risk. Results confirmed that these features of images have an important effect on parents’ attention to the message and their appraisal of risk. This is an important finding, given that appraisal of risk has been shown to affect parents’ implementation of safety precautions.15 ,19

Depicting negative consequences of injury increased the effectiveness of the images, leading parents to rate them as a better fit to the safety message, communicating a greater sense of danger for the child, being more emotionally arousing (greater fear and/or anxiety) and more effectively capturing parents’ attention. Research has also noted that using images that elicit an emotional response promotes attention to the message, and also that parents are more likely to remember these messages and therefore change their behaviour.15 Hence, creating visual images that are consistent with the messages and have some emotional connection to the intended audience seems to maximise the impact of the safety messages on parents’ attitudes and behaviour.17

Limitations and future directions

Although this study provides some significant insights into features of images that are important when directing safety information to parents of young children, it has limitations that should be considered in planning future research. First, the sample was relatively homogeneous, comprising many Caucasian, upper-income and fairly well-educated parents. Tailoring messages to specific sample characteristics can enhance effectiveness of interventions,4 ,20 ,21 and this probably applies also to the images accompanying those messages. Hence, extending this research to examine how features of images affect parents’ appraisals of risk in more diverse populations is an important next step.

Second, the sample comprised only mothers. Gender may influence perceived hazardousness.12 ,22 Some studies find virtually no differences between mothers and fathers in safety practices for young children, whereas others have reported that father–child dyads behave differently from mother–child dyads.23 ,24 Hence, determining if the same results are obtained with fathers is important.

Finally, the features we selected for investigation were chosen based on findings from past research. Although these proved important, there may be other features that are equally, or more, important. Further research may provide additional insights into features of images that enhance messaging impact.

What is already known on the subject

  • Parents’ child-safety practices are influenced by injury-risk appraisals.

  • Images in safety messaging influence these appraisals. Hence, images are important aspects of safety messaging campaigns.

  • Few studies have examined what features of images are important for influencing parent appraisal of injury risk in safety messaging campaigns.

What this study adds

  • Our study evaluates features in images used in the A Million Messages programme, a popular home-safety promotion programme in Canada.

  • The results showed that images that depict injury risk and show injury consequences are rated by parents as having a high impact. Thus, these are important features to include in images associated with home-safety promotion programmes targeting parents.

Speeding trucks targeted

Digital speed monitors have been placed on various streets and roads in at least one district in Australia (Parkes Shire) to prompt truck drivers to keep to the limit. The speed monitors include a radar device to check the speed of passing traffic. Drivers are assured that the speed checks are only a courtesy to remind them to adjust their speed if necessary. Editor’s comment: ‘I cannot believe that this sort of “courtesy” is warranted; this is a situation that demands a tough, consistent response’.


The authors thank the parents for their participation, the members of the A Million Messages (AMM) in Ontario coalition for their support and encouragement and the AMM organisation in Alberta for sharing their images and granting permission for this research and publication. Reprint requests should be addressed to the first author at the Psychology Department, University of Guelph, Guelph, ON N1G 2W1 Canada or



  • Correction Figure 1 of this article has been changed to black and white since published Online First.

  • Contributors BAM is the primary author of this study. She conceptualised the study, developed the methodology and analysis plan and prepared the manuscript for publication. MBe supervised implementation of the study and MBu prepared the modified images and assisted MBe with collection of the data. AK performed the statistical analyses and assisted in preparing the manuscript.

  • Funding This work was supported by Canadian Institutes of Health research grant No 450063 and Social Sciences and Humanities Research Council grant No 430082.

  • Competing interests None.

  • Ethics approval Research ethics board at the University of Guelph approved the project.

  • Provenance and peer review Not commissioned; externally peer reviewed.