Image database of Japanese food samples with nutrition information

Introduction

Visual processing of food plays an important role in identifying inherent biologically-significant information from food, such as edibility (Tsourides et al., 2016). The sight of food elicits affective or hedonic responses (Rodríguez et al., 2005) that occur rapidly even before the conscious perception of food (Sato et al., 2016), which in turn motivates food intake (Yeomans, Blundell & Leshem, 2004). Functional neuroimaging studies have shown that seeing food images and consuming taste solutions activate common neural circuits, including the gustatory and affective brain regions (Simmons, Martin & Barsalou, 2005; for reviews, see Huerta et al., 2014; Van Der Laan et al., 2011).

Studies have created standardized image databases of food to investigate visual food processing (Foroni et al., 2013; Blechert et al., 2014; Miccoli et al., 2014; Charbonnier et al., 2016). For example, Foroni et al. (2013) collected 252 images of natural or processed food, 43 images of rotten food and 529 images of non-food materials in a web-based search and validated these stimuli by assessing the participants’ appraisals, including valence, arousal and familiarity. Their database was used in subsequent studies, such as a functional neuroimaging investigations of visual food processing (Mengotti, Foroni & Rumiati, 2019; Morys, Bode & Horstmann, 2018; Padulo et al., 2016).

However, two issues regarding a food image database remain. First, most of the existing databases contain only Western food. A number of cross-cultural psychological studies have shown that cultural aspects of food modulate hedonic appraisals while seeing and eating food (Wanich et al., 2018; Torrico et al., 2019; Sato et al., 2019b; for a review, see Prescott et al., 1998). For example, Torrico et al. (2019) showed images of various food products to participants with Western and Asian backgrounds and asked them to rate their preferences. The food images from Western and Asian origins elicited stronger preferences in the Western and Asian groups, respectively. These data suggest the need to develop datasets of non-Western food images. To overcome this problem, a recent study developed a food image database containing 209 images in the Asian food category (Toet et al., 2019). It is desirable to develop non-Western food image databases that have different specific advantages to investigate visual food processing further.

Second, the existing databases do not provide nutrition information on the foods. Although the total calories (Foroni et al., 2013; Charbonnier et al., 2016) or the calories and carbohydrate, fat and protein contents (Blechert et al., 2014) in the materials were reported, the information was estimated using general data and not the specific foods in the images (Foroni et al., 2013; Blechert et al., 2014) or are not described in detail (Charbonnier et al., 2016; Toet et al., 2019). This issue could be important, because it is widely believed that objective nutrition information affects the visual processing of food (Birch, 1999) but empirical data are scarce and mixed. Some studies provided positive evidence that participants accurately estimated the total calories (Foroni et al., 2013; Charbonnier et al., 2016; Brunstrom et al., 2018) and fat content (Toepel et al., 2009) and that they showed a preference for high-calorie food (Brunstrom et al., 2018) based on visual information on of the food materials. Other studies reported problematic estimates of total calories for food images (Carels, Konrad & Harper, 2007; Foroni, Pergola & Rumiati, 2016; Horne et al., 2019). To investigate these relationships further, objective nutrition information is indispensable for food image stimuli.

To investigate these issues, we developed an image database of Japanese food samples (i.e., plastic replicas) (Fig. 1). We photographed the stimuli selected from a database of Japanese food samples that looked realistic, included Japanese food that frequently appears in contemporary Japanese home meals, and have detailed nutrition information (e.g., calorie, carbohydrate, fat and protein contents). To validate the image database, we assessed the participants’ subjective appraisals of their affective (valence, arousal, liking and wanting) appraisals of the images. We also assessed cognitive appraisals to confirm whether food images looked as natural as food (naturalness) and were recognizable (recognizability) and familiar (familiarity). As reference data in relation to these appraisals, we additionally included images of Western food selected from an existing database (Blechert et al., 2014). We also assessed the subjective estimate of nutrition and investigated the relationships between objective nutrition information and the subjective perception of nutrition, as well as affective appraisals.

Materials and Methods

Results

Subjective affective and cognitive appraisals

Figure 2 shows the mean (with SE) subjective affective (valence, arousal, liking and wanting) and cognitive (naturalness, recognizability and familiarity) appraisals for each item. Welch’s t-tests contrasting Japanese vs. Western food images showed significant differences for all measures (t(49) > 5.52, p < 0.001, r > 0.61; Table 3), indicating higher appraisals for Japanese foods than for Western foods, except the arousal appraisal (t(49) = 1.00, p > 0.10, r = 0.14). The data suggest that the images of Japanese food samples in the database elicited stronger food-related positive affective appraisals (positive, liking and wanting), and looked more natural, recognizable, and familiar than the photographs of Western food selected from the existing database (Blechert et al., 2014).

Table 3:

Results of Welch’s t-tests (two-tailed) contrasting subjective appraisals for Japanese vs. Western food images.

Statistic	Valence	Arousal	Liking	Wanting	Naturalness	Recognizability	Familiarity
t	5.76	1.21	5.35	6.45	9.29	8.34	7.95
p	0.003	0.277	0.004	0.001	0.000	0.000	0.000
r	0.94	0.46	0.93	0.94	0.98	0.97	0.96

DOI: 10.7717/peerj.9206/table-3

Note:

Degrees of freedom were 4.42–5.87. Significant results (p < 0.05) are in bold.

Relationships between objective nutrition information and subjective food appraisals

Next, we analyzed the relationships between the objective nutrition information (calories, carbohydrate, fat and protein) and subjective nutrition/affective appraisals for food. Figure 3 shows the correlation coefficients between the objective information and subjective estimates of nutrition (cf. group-mean scatterplots with regression lines in Fig. S3). Student’s one-sample t-tests after Fisher’s r-to-z transformation showed that all of the relationships were positive and significant (t(31) > 11.32, p < 0.001, r > 0.53; Table 4).

Table 4:

Results of Student’s one-sample t-tests (two-tailed) for the correlation coefficients between the objective information and subjective estimation of nutrition.

Statistic	Calories	Carbohydrate	Fat	Protein
t	25.62	12.72	11.32	15.04
p	0.000	0.000	0.000	0.000
r	0.98	0.91	0.89	0.93

DOI: 10.7717/peerj.9206/table-4

Note:

Data were analyzed after Fisher’s r-to-z transformation. Degrees of freedom were 31. Significant results (p < 0.05) are in bold.

Figure 4 shows the correlation coefficients between the objective information and affective appraisals (cf. group-mean scatterplots with regression lines in Fig. S4). Student’s one-sample t-tests after Fisher’s r-to-z transformation for objective information showed that objective calories and fat were significantly and positively correlated with all of the affective food appraisals (t(31) > 2.70, p < 0.05, r > 0.37; Table 5). Significant negative correlations were detected between objective carbohydrate and valence, liking, and wanting (t(31) > 2.72, p < 0.05, r > 0.42) and between objective protein and arousal (t(31) = 6.02, p < 0.001, r = 0.69).

Table 5:

Results of Student’s one-sample t-tests (two-tailed) for the correlation coefficients between the objective/subjective nutrition and affective appraisals.

Nutrition	Statistic	Affective appraisal
		Valence	Arousal	Liking	Wanting
Objective calorie	t	2.34	9.28	6.02	3.46
	p	0.025	0.000	0.000	0.001
	r	0.37	0.85	0.72	0.51
Objective carbohydrate	t	2.62	0.98	2.67	2.28
	p	0.013	0.336	0.012	0.029
	r	0.41	0.16	0.42	0.36
Objective fat	t	2.99	6.03	5.05	3.29
	p	0.005	0.000	0.000	0.002
	r	0.46	0.72	0.65	0.49
Objective protein	t	0.84	5.53	−1.11	0.27
	p	0.409	0.000	0.274	0.793
	r	0.14	0.69	0.19	0.05
Subjective calories	t	2.73	9.03	6.73	4.29
	p	0.010	0.000	0.000	0.000
	r	0.42	0.84	0.76	0.59
Subjective carbohydrate	t	0.26	2.27	1.03	0.75
	p	0.799	0.030	0.313	0.459
	r	0.04	0.36	0.17	0.13
Subjective fat	t	1.57	6.06	4.80	3.24
	p	0.127	0.000	0.000	0.003
	r	0.26	0.72	0.64	0.49
Subjective protein	t	0.96	5.98	3.44	2.10
	p	0.342	0.000	0.002	0.044
	r	0.16	0.72	0.51	0.34

DOI: 10.7717/peerj.9206/table-5

Note:

Data were analyzed after Fisher’s r-to-z transformation. Degrees of freedom were 31. Significant results (p < 0.05) are in bold.

In addition, we exploratorily analyzed the relationships between the subjective estimates of nutrition information and affective appraisals (Fig. 4). The analysis showed the same significant patterns as the results of objective information for subjective calories and fat (t(31) > 2.70, p < 0.05, r > 0.37; Table 5), except that the subjective fat–valence relationship did not reach significance (t(31) = 1.57, p > 0.10, r = 0.26). Subjective carbohydrate and protein showed rather different patterns, including a significant positive correlation between subjective carbohydrate and arousal (t(31) = 2.27, p < 0.05, r = 0.36) and negative correlations between subjective protein and arousal, liking, and wanting (t(31) > 2.09, p < 0.05, r > 0.33).

Discussion

Our data on the subjective affective appraisals showed that the images of Japanese food samples were more positive, liked, and wanted by Japanese participants than the five selected images of Western food items. The naturalness appraisals showed that the images of Japanese food samples were rated more natural than the Western food images selected from the existing database. This result suggests that the food sample images looked almost as real as actual food. The results for familiarity and recognizability showed that the Japanese foods in the database were more familiar and more easily recognizable by Japanese participants than the selected Western food items. Overall, these results indicate that the images of Japanese food samples we developed validly represent Japanese foods.

Because our food image database had objective nutrition information, we analyzed the relationships between objective nutrient information and subjective nutrition and affective appraisals. The results showed that the objective information and subjective perception of nutrition were positively associated, and that the objective caloric content and percentage of fat calories were positively associated with all of the affective appraisals. These results are consistent with previous findings that participants estimated the caloric (Foroni et al., 2013) and fat (Toepel et al., 2009) contents precisely and reported a greater liking for food with high caloric values (Brunstrom et al., 2018) during visual food processing, although other studies failed to find such patterns (Carels, Konrad & Harper, 2007) and debate remains. However, previous studies estimated the nutrition information using a general database and did not investigate the contents of specific food images. The previous studies also did not investigate systematically the carbohydrate, fat, and protein contents. Therefore, we investigated this and found that the relative caloric ratio of carbohydrate and protein was negatively associated with the affective appraisals of food. Our results also showed that the relationships between objective nutrition information and affective appraisals are rather similar to those between subjective nutrition perception and affective appraisals in terms of caloric and fat contents, but not in terms of carbohydrate and protein. In summary, our results confirm and extend previous findings indicating that objective nutrition information about food can influence the subjective estimates of nutrition and affective appraisals during the visual processing of food.

The database of Japanese food images with nutrition information that we developed has practical significance for research on visual food processing. For example, it would be interesting to use our database in functional neuroimaging studies. Previous studies have shown that several brain regions, including the visual cortices (e.g., the fusiform gyrus) and limbic regions (e.g., the amygdala), are activated more during the presentation of food images than non-food images (Holsen et al., 2005; Sato et al., 2019a; for a review, see Van Meer et al., 2015). Although a few studies compared neural activity in response to low vs high calorie/fat food (Frank et al., 2010; Killgore et al., 2003; Toepel et al., 2009), more detailed parametric relationships between neural activity and objective nutrition information remain unexplored. Use of the current database may allow studies of the specific brain activities associated with objective nutrition.

Several limitations of this study should be acknowledged. First, our database was restricted to a small number of food samples. Therefore, the number of images should be increased. Developing a comparable image database using real food with detailed nutrition information may also be helpful. Second, our stimuli were restricted to processed food. Because some previous studies have suggested different psychological (Aiello et al., 2018; Coricelli et al., 2019a; Rumiati et al., 2016) and neural (Coricelli et al., 2019b; Pergola et al., 2017) processing between processed and non-processed food, the generalizability of the current results for non-processed food is an important matter for future research. Third, our food stimuli were depicted in different plates. This was because the Japanese diet has custom rules for plates (Thompson, 2016) and the food sample set we photographed mimicked this to create realistic Japanese food stimuli. This may be problematic, because some previous studies have shown that plates can influence the affective response to food items (Piqueras-Fiszman et al., 2012; Piqueras-Fiszman, Giboreau & Spence, 2013; Stewart & Goss, 2013; Van Ittersum & Wansink, 2012). An assessment using the present images after removing the plates is warranted to confirm the findings. Fourth, we assessed only subjective appraisals. Because subjective appraisals could be biased to the results that researchers want to find due to participants’ care about demand characteristics (Orne, 1962), objective measures of affective responses, such as physiological signals (Kaneko et al., 2018), may complement the current findings. Finally, we tested only a small sample of participants. We determined the sample size to detect more than middle size effects (cf. Cohen, 1992). Because our sample included only young participants, it remains unknown whether participants of different age groups, such as children and older participants, would show similar affective and cognitive appraisals, and concordance between objective nutrition information and subjective appraisals. We tested only Japanese participants; hence, the patterns may be different in different cultures. We also did not assess the details of participants’ characteristics that could modulate food processing, such as dieting habits (Coricelli et al., 2019a; Hoefling & Strack, 2008). Investigations including more participants from different age groups and different cultures, with detailed assessments of their characteristics, by using the present stimulus database would be valuable for investigating visual food processing further.

Conclusions

We developed an image database of Japanese food samples that contains detailed nutrition information, including calorie, carbohydrate, fat and protein contents. The appraisal experiments showed that all affective and cognitive appraisals of the Japanese food sample images were higher than, or comparable to, those of Western foods. Correlational analyses showed positive associations between the objective information and subjective perception of the calorie, carbohydrate, fat and protein contents, and objective calorie/fat content and affective appraisals. These data suggest that by using our image database, researchers can investigate the visual processing of Japanese food and relationships between objective nutrition information and psychological/neural processing of food. This image database is available on request from the corresponding author solely for research purposes. The nutrition information for the stimuli is available from the Iwasaki SAT system (http://www.foodmodel.com/category12/index.html).

Supplemental Information