Remote dining is a way of using information technology to support commensality, allowing people in different places to eat together through video calls. Niewiadomski et al. (2019) reviewed works on computational commensality over the past 5 years and found that researchers paid more attention to health, wellbeing, and food experiences in human-food interactions. Moreover, they divided digital commensality into two kinds according to whether the partners were human. Remote dining involving human-human interactions has become a way to create or maintain connections between individuals via shared technologies. During the COVID-19 pandemic, researchers, governments, and marketing companies have conducted surveys on remote dining and showed the increased use and challenges of remote dining (Ceccaldi et al., 2020; Recruit Co., Ltd, 2020; Ministry of Agriculture, Forestry, and Fisheries, 2021). For instance, a marketing report of a survey of about 10,000 people showed that 18.3% of Japanese people in their 20s−60s who live in urban areas have attended an online party and that online parties are usually attended by three to six people (Recruit Co., Ltd, 2020). Due to increased demand on remote dining, it has been suggested that more research is needed to improve its design and experience (Niewiadomski et al., 2019; Spence et al., 2019; Ceccaldi et al., 2020). For instance, Furukawa and Inoue (2013) conducted an experiment involving both face-to-face and remote co-eating and found that meal visibility had positive effects on conversations and impressions of co-eating experience. Konno et al. (2017) conducted a two-month experiment involving two families and suggested that remote dining involving elderly parents and their children can increase parents' meal satisfaction. Because of the barriers to co-eating caused by time zone differences (Scander et al., 2021), some studies have focused on asynchronous remote co-eating systems and the interactions therein (Noguchi and Inoue, 2016; Obata et al., 2020).

Interpersonal synchronization occurs when people are together because of maintained engagement, temporal coordination, or contingency (Harrist and Waugh, 2002). In co-eating situations, eating behavior has been found to be affected by the partners because of “social facilitation”, “social inhibition” or “social modeling” (De Castro, 1990, 1994; Lumeng and Hillman, 2007; Hartmann et al., 2010; Cruwys et al., 2015; Herman, 2015; Kaisari and Higgs, 2015; Nakata and Kawai, 2017; Kawai et al., 2021; Kimura et al., 2021). From a broad perspective, people coordinate their eating pace, especially when sharing food, either eating with friends or strangers (Woolley and Fishbach, 2019). Noguchi and Inoue (2016) suggested that coordinating the eating pace of recorded videos for co-eating had effects on the conversation in its speech frequency and length of switching pauses. Woolley and Fishbach (2019) found a positive relationship between the subjective feelings of coordination of meals and trust. From a narrower perspective, the mimicry of food intake has also been studied during meals either with family members or strangers (Hermans et al., 2012; Sharps et al., 2015; Bell et al., 2019). These studies have highlighted the likelihood of biting mimicry, which refers to the likelihood of people eating when others eat and not eating when others do not. Certain situational factors that may influence mimicry have also been discussed. Hermans et al. (2012) found that the likelihood of biting mimicry was higher in the first half of a meal. Sharps et al. (2015) studied mimicry between children and parents when multiple meal items were served freely on separate plates and found that children's biting mimicry occurred only when both they and their parents were eating the same items. Bell et al. (2019) investigated possible distractions in real-life environments that can lead to reduced mimicry.

Individuals may pay attention to what others eat in co-eating scenarios and usually compare their eating choices and amounts to those of others (Cruwys et al., 2015). Food is considered an individual identity element used to determine whether people share a culture or beliefs (Fischler, 1988; Polivy, 2017). Choosing similar types of food is considered a way to cater to each other (Cavazza et al., 2017). Some studies have investigated the effect of meal similarity on impressions of the partners. Woolley and Fishbach (2019) conducted a series of experiments to explore the influence of eating similar food on trust and found that participants who were assigned similar snacks trusted each other more. Likewise, participants also trusted advertised product information more when the advertiser ate a similar snack. It should be aware of that the participants in that study did not interact with each other but only observed each other's eating behavior. Furthermore, in face-to-face situations, people may exchange food or use shared plates. Sharing food can increase coordination and promote bonding, thus implying a positive social relationship (Miller et al., 1998; Fischler, 2011; Woolley and Fishbach, 2019). Besides, Kaisari and Higgs (2015) explored the effects of meal similarity on social modeling of food intake and found that modeling occurred either eating the same or different snacks in the meal.

Thirty-six women aged 22 to 58 years (M = 41, SD = 11.2 years) participated in the study voluntarily and were divided into twelve groups (G1–G12). In this study, we recruited adults with jobs, which were major users of online catering services. Considering the COVID-19 pandemic, we recruited participants with no sick symptoms and excluded people working in the medical and welfare industries. The participants had no health issues and had no problems having a meal and a conversation. Consistent with previous research, all the participants were women (Mukawa et al., 2009; Hermans et al., 2012; Sharps et al., 2015), and all the groups consisted of strangers (Woolley and Fishbach, 2017) to avoid the influences of mixed genders and different relationships. The participants were recruited through a recruitment agency and grouped according to their time availability. Each participant received 6,500 yen for 90 min of participation.

We adopted a within-subjects design to avoid the effects of personal eating habits, as some people are more talkative than others during a meal or can eat more comfortably in the presence of others. To explore the effects of meal items, triads of participants were provided with “same meal items,” “different meal items,” and “no meal items” (only drinks) lunch boxes. The participants were exposed to these three conditions in three remote sessions with the same partners.

In each session, the participants chatted freely via a video call lasting 15 min. Chatting topics were not assigned. The participants generally talked about their daily lives, upcoming events, and sometimes food and cooking. In the “same meal item” and “different meal item” conditions, each participant was provided with a lunch box containing nine small dishes, a rice dish, and a cup of tea to consume while chatting. Before each session, the menu was shown to the participants. This included the lunch box that each participant was offered, and the name and picture of each dish. The participants took a glance at the menu but did not memorize it so that they could still ask their partners what food they were having. The eating pace and amount were not predetermined, and no participants had finished their meals by the 8 min of dining.

Regarding the experimental environment, the participants in each group were shown to three different rooms in Kasuga Area, University of Tsukuba. The participants were seated in a chair facing a 42-inch display (Panasonic TH-L42E60) 100 cm away. A laptop PC was connected to the display, and Zoom was used for the video conferences (25 fps). A USB webcam with a 1920 × 1080 resolution was used, and each partner's face and meal items could be clearly seen on the screen. Each participant's own image was displayed at the center, while those of the partners were shown on the left and right above, like eating at a round table. The built-in speakers of the display and the built-in microphone of the webcam were used, and the volume of the partners' voices was adjusted to a normal conversational level. The screens during the sessions were recorded by Zoom for video analysis. Another two cameras in each room, one in front of the participant and the other behind the entire environment, recorded backup videos. One experimenter was behind the curtain in each room during the sessions and set up the lunch box and questionnaire between the sessions.

Four lunch box sets (A–D) sold for online dining parties were provided by online catering company nonpi Inc. (2021). We used four menus, called “Standard Plan Japanese,” “Standard Plan Western,” “Miyabi Early Summer,” and “Miyabi Taste of Edo,” in which the amounts of food were similar, but the style, ingredients, and cooking methods were different (Table 1). To compose four lunch box sets with no duplicate dishes, we used nine small dishes and a rice dish with a cup of cold oolong tea. The rice dishes were a bowl of rice with salmon, a bowl of rice with tomato, a bowl of onigiri, and three onigiri wrapped in leaves. To avoid the effects of lunch box set type, the use of sets A–D was balanced, with each set used in the “same meal items” condition for three random groups and in the “different meal items” condition for the other nine groups.

To determine the features of the four lunch box sets, we identified the main ingredients, size, and hardness of the nine dishes in each set. First, we identified the main ingredients according to the photograph and menu name of each dish. We then identified the sizes of the main ingredients according to the photograph. We used a three-centimeter cube (Koga et al., 2016) as a standard to determine whether the main ingredients could be eaten in one bite and divided them into three levels: small (all sides <3 cm) = 1, long (one side >3 cm) = 2, and big (two sides >3 cm) = 3. We expected that the larger the food was, the more likely the participants were to cut it with chopsticks, which might affect the food holding duration. Lastly, we determined hardness according to a hardness list of common ingredients and cooking methods compiled by Sekiguchi et al. (1996) using a tactile sensor. If there were no directly corresponding ingredients on the list, we used similar ingredients to estimate their hardness. We divided hardness according to the eight levels on the list. Typically, levels 1–2 include soft dishes, such as tofu (bean curd) and fish; levels 3–4 include hard dishes, such as vegetables and meat; and level 5 includes nuts and fried food. Levels 6–8 include snacks, which were not used in this study. It is considered that, given the same food size, the harder the food, the more chewing is required (Michiwaki et al., 2001).

The features of the four lunch box sets determined according to the aforementioned method are shown in Table 1. The average sizes of the dishes in lunch boxes A–D were 2.1, 2.2, 2.7, and 3, respectively. The average hardness levels of lunch boxes A–D were 2.7, 3.1, 2.2, and 2.9, respectively. In lunch box A, there were some soft dishes, such as fish and tofu, as well as some hard dishes, such as nuts and vegetables. In lunch box B, there were more hard dishes, such as nuts and meat. In lunch box C, there were more soft dishes, such as fish and tofu. In lunch box D, there were more hard dishes, such as fried foods.

The experiment consisted of three sessions and lasted about 90 min in total. In the experiment, the participants got with the same person 3 times. To minimize the order effects such as feeling less hungry or getting familiar with the partners in the later sessions, the three sessions with the three conditions were balanced with six different orders, which were S → D → N, S → N → D, D → S → N, D → N → S, N → S → D, and N → D → S (S: same meal items, D: different meal items, N: no meal items). Each of the orders was assigned to two groups of participants randomly. The following procedure was followed (the order of S → D → N is used as an example):

We used video recordings to analyze each participant's eating behavior. We focused on 3–8 min of each 15-min remote dining session, because the participants first greeted each other and opened their lunch boxes and because interpersonal synchronization is more likely to occur in the first half of a meal (Hermans et al., 2012). The backup videos recorded by the cameras in front of and behind each participant were occasionally used to supplement the video call recordings.

The video recordings of each participant (6 h in total) were manually labeled using ELAN (Figure 1), a video analysis tool, and analyzed quantitatively by two authors. The labels were predefined in a codebook by both observers. The two observers labeled a 5-min video of one group and resolved any disagreements after discussion before labeling other videos. Codes for actions that had not been anticipated, such as licking chopsticks, were decided through discussion and added to the codebook. There were no more code additions after half of the groups had been labeled. After labeling, we calculated and compared the proportions of the total duration of participants' eating actions in 5 min (%), frequency of participants' eating actions (times per minute), and average length of participants' eating actions (seconds per time) in the “same meal items” and “different meal items” conditions.

According to previous studies, the process of food consumption includes holding tableware, holding food, taking a bite (putting food in the mouth), and chewing (Mukawa et al., 2009; Inoue et al., 2019). In this study, we labeled and analyzed the durations of holding food and biting, followed by chewing, as the main elements of eating behavior. We considered putting food in the mouth to be a kind of biting, so these two actions were treated as one. We also labeled the drinking period and the period of not holding tableware as resting periods.

To examine the participants' subjective feelings, we administered a questionnaire rated on a seven-point Likert scale (where one represented strongly disagree, and seven represented strongly agree) with a section for free comments. This was followed by a short interview. The questionnaire used in the experiment consisted of forty-eight items. Twenty-one of these items were analyzed in this study to focus on impressions of the tele-dining experience. Other twenty-seven items were related to intentions and perceptions of conversations, intentions and perceptions of gaze directions, environments of the experiment, and were excluded from the analysis because they were out of focus on the topic of this article. Among the twenty-one items analyzed in this study, eight items were selected from the studies focusing on impressions of co-eating experiences (Furukawa and Inoue, 2013), three of them were selected from the studies focusing on closeness of co-eating partners (Woolley and Fishbach, 2017), and necessary modifications were made for clarity. Other ten items were developed with reference to (Ceccaldi et al., 2020) to explore whether people's expectations of co-eating were satisfied, such as “feeling the food is delicious” and “feeling like eating together”. The items used in this study fell under the following eight categories:

To answer the research questions, we used non-parametric analysis as some of the items in eating actions, interpersonal synchronization and subjective feelings were not normally distributed according to Shapiro-Wilk normality test. To answer RQ1, RQ2 and RQ3, Wilcoxon rank-sum test was used to compare the eating actions, interpersonal synchronization and subjective feelings between the “same meal items” and “different meal items” conditions, as the data were paired. In addition, Kruskal-Wallis test was used to compare the eating actions between the participants eating lunch box set “A”, “B”, “C” and “D”, as the data were not paired. Lastly, to answer RQ4, we used Spearman's rank correlation coefficient to investigate the relationship between either-two-interpersonal synchronization and subjective feelings.

The median of participants' levels of hunger were 5.5 and 4 before they ate their first and second meals respectively on a seven-point Likert scale (where 1 represented strongly disagree with “I am hungry.”, and 7 represented strongly agree with “I am hungry.”). Overall, the participants were moderately hungry when they participated in the first session with a meal and were neutral when they participated in the second session with a meal.

The proportions of the total duration of participants' eating actions in five minutes (%), frequency of participants' eating actions (times per minute), and average length of participants' eating actions (seconds per time) are shown in Figure 4. Overall, the median of proportions of holding food in the “same meal items” and “different meal items” conditions were 36 and 37%, respectively. The mean proportions of chewing in the “same meal items” and “different meal items” conditions were 54 and 47%, respectively. To compare the time spent holding food and chewing between the conditions (Figure 4A), the Wilcoxon rank-sum test showed that proportions of chewing differed significantly between the conditions (N = 36, W = 200, p = 0.0373). This suggested that the participants spent more time eating when they were having the same food.

Regarding the eating pace, the participants held food 3.6 times and took 2.7 bites per minute in the “same meal items” condition, whereas they held food 2.8 times and took 2.0 bites per minute in the “different meal items” condition. The median lengths of holding food and holding food per bite were 6.0 and 7.5 sec, respectively, in the “same meal items” condition and 7.2 and 9.2 sec, respectively, in the “different meal items” condition. The median length of chewing was 11.4 sec in the “same meal items” condition and 12.4 sec in the “different meal items” condition. To comparing the eating pace between the conditions (Figures 4B,C), the Wilcoxon rank-sum test showed that the frequencies of holding food and biting differed significantly between the conditions (N_Hf = 34, W_Hf = 158.5, p_Hf = 0.018; N_B = 34, W_B = 158.5, p_B = 0.018). The median lengths of holding food and holding food per bite showed marginally significant and significant differences, respectively, between the conditions (N_Hf = 36, W_Hf = 214, p_Hf = 0.063; N_HfpB = 36, W_HfpB = 200, p_HfpB = 0.037). This suggested that the participants ate at a faster pace when they were having the same food.

Regarding the resting periods, in both conditions, 18 participants drunk the water and the median proportion of drinking of these people is 1%. In the “same meal items” and “different meal items” conditions, 15 participants and 23 participants put down their tableware and the median proportion of not holding tableware were 3 and 11% of these people, respectively. The Wilcoxon rank-sum test showed that the proportions of not holding tableware of the whole participants differed significantly between the conditions. This suggested that the participants put their tableware down less frequently when they were having the same food.

The numbers of pairwise synchronized bites of each pair in 5 min and the proportions of pairwise synchronized biting in each pair are shown in Figure 5. The medians of pairwise synchronized bites were 3 and 2 in the “same meal items” and “different meal items” conditions, respectively. The median proportions of pairwise synchronized biting were 25 and 20%, respectively. To compare the instances and proportions of pairwise interpersonal synchronization, the Wilcoxon rank-sum test showed that the instances and proportions of pairwise synchronized biting differed significantly between the conditions (N _numbers = 60, W_numbers = 520, p_numbers = 0.003; N_proportion = 69, W_proportion = 834, p_proportion = 0.026; Figures 5A,B). In terms of the likelihood of biting mimicry, the ratios of sensitive to nonsensitive bites did not differ significantly between the conditions: 1.10 and 0.92 in the “same meal items” and “different meal items” conditions (Figure 5C), respectively. This suggested that the tendency “to eat when partners eat and not to eat when partners do not” did not differ between the two conditions.

The numbers of triad synchronized bites of each person in five minutes and the proportions of triad synchronized biting of each person are shown in Figure 6. The total of triad synchronized bites were 31 and 15, and the medians were 1 and 0 in the “same meal items” and “different meal items” conditions, respectively. The median proportions of triad synchronized biting were 5 and 0%, respectively. To compare the instances and proportions of triad interpersonal synchronization, the Wilcoxon rank-sum test showed that the instances of triad synchronized biting marginally differed significantly between the conditions and proportions of triad synchronized biting differed significantly between the conditions (N _numbers = 22, W_numbers = 72.5, p_numbers = 0.075; N_proportion = 24, W_proportion = 78, p_proportion = 0.041; Figures 6A,B).

Lastly, the numbers of either-two-synchronized and either-two-passive synchronized bites of each person in 5 min, the proportions of either-two-synchronized and either-two-passive synchronized biting of each person, are also shown in Figure 6. The medians of either-two-synchronized bites were both 5 in the “same meal items” and “different meal items” conditions, respectively. The median proportions of either-two-synchronized biting, representing the proportions of the biting that following any partner's biting, were 48 and 42%, respectively. The medians of either-two-passive synchronized bites were 6 and 5, respectively. The median proportions of either-two-passive synchronized biting, representing the proportions of the biting that is followed by any partner's biting, were 39 and 36%, respectively. To compare the instances and proportions of either-two-interpersonal synchronization, the Wilcoxon rank-sum test showed that the instances of either-two-synchronized biting and either-two-passive-synchronized biting differed significantly between the conditions (N_{either−two−synchronization} = 29, W _{either−two−synchronization} = 112.5, p _{either−two−synchronization} = 0.023; N _{either−two−passivesynchronization} = 31, W _{either−two−passivesynchronization} = 126, p _{either−two−passivesynchronization} = 0.017; Figures 6C,E). The median of the difference of the number of either-two-synchronized and either-two-passive- synchronized biting between two conditions (same meal item condition – different meal item condition) were 1 and 1.5. This suggested that there were more instances of either-two-interpersonal synchronization when the participants were having the same food. However, the proportion of either-two-synchronized or either-two-passive synchronized biting did not differ significantly between the conditions.

The mean lengths of holding food per bite and chewing (seconds per time) are shown in Figure 7. To compare the mean lengths of holding food per bite and chewing between participants eating each lunch box set, we used the Kruskal-Wallis test, as some items were not normally distributed. There were no statistically significant differences between participants eating each lunch box set. The distribution of chewing duration using the four lunch box sets is shown in Figure 8.

The average scores of subjective feelings in each category are shown in Figure 9.

In terms of tastiness, the participants generally reported liking the food (Md_S = 6; Md_D = 5). There was a marginally significant difference between the conditions (N = 15, W = 29.5, p = 0.0775), suggesting that the participants found the food more likable when they were having the same meal.

Regarding their feelings about the meal, the participants had a generally positive attitude toward “enjoyment of meal” (Md_S = 5.3; Md_D = 5.0; Cronbach's alpha = 0.68) but a neutral attitude toward “ease of meal” (Md_S = 4.3; Md_D = 4.2; Cronbach's alpha = 0.58). In terms of feelings about the conversation, the participants had a generally positive attitude toward “enjoyment of conversation” (Md_S = 4.8; Md_D = 5.0; Cronbach's alpha = 0.68) and “ease of conversation” (Md_S = 5.0; Md_D = 4.8; Cronbach's alpha = 0.58).

Regarding their feelings about companionship, the participants had a generally positive attitude toward “having a meal together” (Md_S = 5.0; Md_D = 5.0; Cronbach's alpha = 0.81) and “closeness” (Md_S = 5.0; Md_D = 5.0; Cronbach's alpha = 0.93) but a neutral attitude toward “food topics” (Md_S = 3.5; Md_D = 3.3; Cronbach's alpha = 0.54). Among the items in the “food topics” category, the participants reported having conversations about the “taste of food” (Md_S = 5; Md_D = 5) but not about their “food preferences” (Md_S = 3; Md_D = 2) or “memories of food” (M_S = 2; M_D = 2). The Wilcoxon rank-sum test showed a marginally significant difference in the feeling of “having a meal together” between the conditions (N = 24, W = 89, p = 0.0809), suggesting that the participants had a greater sense of having a meal together when they were having the same meal.

In seventy-two samples collected from thirty-six participants in the two conditions, Spearman's rank correlation coefficient showed that interpersonal synchronization positively correlated with the feeling of “ease of meal”: “ease of meal” modestly correlated with the numbers of either-two-synchronized biting (rho = 0.29, p = 0.011), and moderately correlated with the numbers of either-two-passive synchronized biting (rho = 0.41, p = 0.000). This suggested that interpersonal synchronization was related to the ease of co-eating. Scatterplots of the numbers of synchronized biting and subjective feeling and the numbers of passive synchronized biting and subjective feeling are shown in Figure 10. Surprisingly, no other positive correlations were found.

Besides the abovementioned positive correlation, there was a modestly negative correlation between the numbers of either-two-synchronized biting and the feeling of “enjoyment of conversation” (rho = −0.21, p = 0.082) but not between either-two-passive synchronized biting and “enjoyment of conversation” (rho = −0.10, p = 0.397).