The authors have declared that no competing interests exist.
Conceived and designed the experiments: JQ MT. Performed the experiments: JQ MT. Analyzed the data: DT JQ MT. Contributed reagents/materials/analysis tools: DT JQ MT. Wrote the paper: DT JQ MT. Development of study concept: DT JQ MT.
Current address: DVJ Insights market research, Utrecht, the Netherlands
Despite decades of research establishing the causes and consequences of emotions in the laboratory, we know surprisingly little about emotions in everyday life. We developed a smartphone application that monitored real-time emotions of an exceptionally large (N = 11,000+) and heterogeneous participants sample. People’s everyday life seems profoundly emotional: participants experienced at least one emotion 90% of the time. The most frequent emotion was joy, followed by love and anxiety. People experienced positive emotions 2.5 times more often than negative emotions, but also experienced positive and negative emotions simultaneously relatively frequently. We also characterized the interconnections between people’s emotions using network analysis. This novel approach to emotion research suggests that specific emotions can fall into the following categories 1)
Hundreds of papers in psychology, medicine, marketing, management, and many other fields begin by asserting that emotions are ubiquitous to human life. But exactly how “ubiquitous” are they? A tremendous body of work has established that various stimuli and situations can cause emotions [
Recent years have witnessed an explosion of research on specific emotions. In particular, a fast growing body of work aims to investigate the health benefits of specific emotions such as gratitude [
Beyond frequency, the wide range of emotions people can experience prompts the question of how the main specific emotions are interrelated. For example, is anger more likely to be experienced in tandem with anxiety or with sadness? Can we feel love and contempt at the same time? We know very little about which emotions typically co-occur or are rarely or never experienced in tandem. Existing research on the structure of affect has primarily provided insight into the factors that may underlie emotions. For instance, the best-known model of affect is the circumplex model, which proposes that emotions can be ordered on the circumference of a circle that comprises two orthogonal psychological dimensions: valence and arousal—the distance between two specific emotions corresponds to the similarity and correlations between them [
Several methodological challenges have made studying the frequency and centrality of emotions as they are experienced in everyday life with a large and diverse group of people a particularly difficult endeavour. We sought to overcome these challenges by developing a multiplatform experience sampling smartphone application, which yielded real-time measures among an exceptionally large group of people. This approach allowed us to examine three fundamental questions about human emotions: 1) how often do people experience emotions in general, 2) which emotions do people specifically experience (i.e., frequency), and 3) how central are different emotions within the emotion network (i.e., centrality)?
Participants volunteered for the study by downloading “58 seconds”, a free francophone mobile application for both iPhone and Android dedicated to measuring various aspects of users’ psychological experience through short questionnaires presented at random times throughout the day. The project received significant coverage on national television in both France and Belgium, totalizing over 60,000 users and half a million questionnaires completed.
We developed an emotion questionnaire that was embedded in a larger study. Specifically, participants were asked to indicate whether or not they were currently feeling nine specific positive (alertness, amusement, awe, gratitude, hope, joy, love, pride, and satisfaction) and nine specific negative (anger, anxiety, contempt, disgust, embarrassment, fear, guilt, offense, and sadness) emotional states, which were adapted from the modified Differential Emotion Scale (mDES) (mDES: [
The present sample represents over a year of data collection (February 2013 to April 2014). Note that the application is still running and more data are currently being collected. In total, 11,572 participants (Mage = 32.9, SDage = 10.4; 75% women) completed a total number of 65,721 emotion reports over an average of 35 days (median = 8.6).
75% female (N = 49,039), 25% male N = 16,682) | |
93% French; %% Swiss; 2% other; 0,5% Belgian | |
Number of questionnaires completed during different days of the week and times of the day.
Participants were given no financial compensation but were provided once a week with feedback on their aggregated levels of emotions. Participants answered several demographic questions, and were asked which days of the week and within what time windows they wished to receive questionnaire requests (default = 7 days/week from 9 AM to 10 PM). Participants could also customize the number of daily questionnaire requests they wished to receive (default = 4, minimum = 1, maximum = 12). The application algorithm then chose random times to send questionnaires within each participant’s day, with a minimum of one hour between two questionnaire requests. The random sampling was ensured through a notification system that did not require users to be connected to the Internet. New random times were generated each day, and the times were independently randomized for each participant. Thus, although participants were free to choose the time window in which they received questionnaires, the exact times they received a questionnaire was randomly determined, with at least one hour between two questionnaires. At each of these times, participants received a notification on their mobile phone informing them that a new questionnaire was available. They then had the possibility to take the questionnaire, snooze it (i.e., delay the questionnaire request by 9 minutes), or reject it.
We attempted to avoid some of the pitfalls of potential self-selection bias posed by smartphone research, including a potentially younger and wealthier sample than the average population, by advertising our research project on various channels—from local newspapers to national primetime television—and by making our experience sampling application a multiplatform one. That is, while other applications are exclusively developed for the relatively expensive iPhone (see e.g.,
The Ethics Committee of the University of Groningen, the Netherlands, approved the study in written form. The study method was carried out in accordance with the approved guidelines. All study protocols were approved by the aforementioned Committee. At initial signup, participants provided their written informed consent. The data have been deposited on Open Science Framework (
To take into account the nested structure of our data (most participants answered multiple times), estimations of the frequency of emotions were obtained using multi-level modelling with a compound symmetry covariance matrix. We computed the percentage of the time that people reported feeling an emotion; that is, the frequency of moments for which participants indicated they experienced at least one of the 18 emotions on the list. On average, people reported experiencing one or several emotions 90% of the time. Specifically, participants indicated experiencing one or several positive emotions (i.e., positive emotions only) 41% of the time, one or several negative emotions (i.e., negative emotions only) 16% of the time, and at least one positive and one negative emotion simultaneously (i.e., mixed emotions) 33% of the time. Breaking down these results by emotion, we computed the frequency of moments for which participants indicated they experienced each of the 18 distinct emotions on the list. As depicted in
Frequency | Centrality | |||
---|---|---|---|---|
Emotion | Percentage | 95% CI Lower Bound | 95% CI Upper Bound | Degree Centrality |
Joy | 35 | 34 | 35,2 | 3 |
Love | 30 | 29,3 | 30,6 | 1,4 |
Anxiety | 29 | 28,6 | 29,8 | 1,7 |
Satisfaction | 27 | 26,5 | 27,6 | 2,7 |
Alertness | 24 | 24 | 25 | 1,3 |
Hope | 22 | 21,8 | 22,9 | 2 |
Sadness | 20 | 19,7 | 20,8 | 2,6 |
Amusement | 16 | 15,9 | 16,8 | 1,8 |
Pride | 13 | 12,7 | 13,6 | 2,2 |
Disgust | 11 | 11 | 11,8 | 2,3 |
Anger | 10 | 9,5 | 10,2 | 2,2 |
Gratitude | 9 | 8,6 | 9,4 | 2,1 |
Guilt | 5 | 5,2 | 5,7 | 1,6 |
Fear | 5 | 5,1 | 5,7 | 1,6 |
Awe | 5 | 4,9 | 5,5 | 1,9 |
Offense | 5 | 4,7 | 5,2 | 1,8 |
Embarrassment | 5 | 4,4 | 4,8 | 1,2 |
Contempt | 1 | 1 | 1,2 | 1 |
Positive emotion only | 41 | 40,1 | 41,4 | |
Negative emotion only | 16 | 15,7 | 16,6 | |
Mixed emotion | 33 | 32,3 | 33,6 | |
ANY EMOTION | 90% | 89,3 | 90,41 |
In order to provide a detailed account of emotion in everyday life, we further broke down our results by reporting the frequency of emotions across the different days of the week and time of the day. Because relatively few people provided emotion reports from 11PM to 5AM (all these measurement times had fewer than 1,000 reports; see
Regarding the experience of specific positive emotions across different times of the day,
The line colors between specific emotions represent the extent to which emotions tend to co-occur (blue hues) or inhibit each other (red hues). The numbers in the grey dots underneath specific emotions represents their frequency of occurrence in the sample. The right panel represents the percentage of times respondents reported experiencing any, positive, negative, or mixed emotions.
Gender differences: In order to explore potential gender differences in the frequency of emotional experience in everyday life, we analyzed the data separately for men and women.
Women | Men | |||||||
---|---|---|---|---|---|---|---|---|
Emotion | Frequency (% time) | 95% CI Lower Bound | 95% CI Upper Bound | Degree Centrality | Frequency (% time) | 95% CI Lower Bound | 95% CI Upper Bound | Degree Centrality |
Joy | 34.02 | 33.36 | 34.67 | 2.92 | 36.07 | 34.86 | 37.28 | 3.03 |
Love | 30.10 | 29.30 | 30.90 | 1.33 | 29.56 | 28.22 | 30.91 | 1.43 |
Anxiety | 30.31 | 29.59 | 31.04 | 1.68 | 26.13 | 24.95 | 27.31 | 1.84 |
Satisfaction | 25.85 | 25.21 | 26.48 | 2.68 | 30.59 | 29.45 | 31.74 | 2.81 |
Alertness | 23.06 | 22.31 | 23.81 | 1.29 | 27.44 | 26.36 | 28.52 | 1.35 |
Hope | 22.34 | 21.70 | 22.99 | 2 | 22.48 | 21.38 | 23.58 | 2.22 |
Sadness | 21.30 | 20.69 | 21.92 | 2.5 | 17.12 | 16.12 | 18.13 | 2.71 |
Amusement | 15.78 | 15.15 | 16.41 | 1.78 | 18.10 | 17.20 | 19.00 | 1.79 |
Pride | 12.23 | 11.66 | 12.81 | 2.16 | 15.73 | 14.77 | 16.69 | 2.36 |
Disgust | 11.68 | 11.23 | 12.13 | 2.27 | 10.46 | 9.69 | 11.22 | 2.49 |
Anger | 10.27 | 9.86 | 10.67 | 2.19 | 8.51 | 7.85 | 9.17 | 2.29 |
Gratitude | 9.18 | 8.73 | 9.62 | 2.07 | 8.33 | 7.41 | 9.25 | 2.18 |
Guilt | 5.43 | 5.12 | 5.74 | 1.53 | 5.57 | 5.00 | 6.13 | 1.81 |
Fear | 5.68 | 5.34 | 6.02 | 1.53 | 4.49 | 3.97 | 5.01 | 1.73 |
Awe | 5.09 | 4.70 | 5.48 | 1.87 | 5.31 | 4.62 | 6.00 | 1.96 |
Offense | 5.19 | 4.90 | 5.49 | 1.75 | 4.13 | 3.69 | 4.58 | 1.8 |
Embarrassment | 4.57 | 4.29 | 4.85 | 1.12 | 4.67 | 4.18 | 5.16 | 1.41 |
Contempt | 0.97 | 0.81 | 1.14 | 0.85 | 1.35 | 1.08 | 1.61 | 1.45 |
Positive only | 38.96 | 38.11 | 39.80 | — | 45.30 | 43.98 | 46.61 | — |
Negative only | 16.83 | 16.30 | 17.35 | — | 14.02 | 13.15 | 14.89 | — |
Mixed emotion | 33.81 | 33.10 | 34.51 | — | 30.51 | 29.32 | 31.70 | — |
ANY EMOTION | 89.92 | 89.38 | 90.47 | — | 89.93 | 89.02 | 90.84 | — |
The aforementioned results tell us how ubiquitous, in terms of frequency, various specific emotions are in people’s daily lives. But emotions might also differ in how they relate to other distinct emotions within the emotional network. In other words, some emotions may typically stimulate or inhibit the experience of other emotions, whereas other emotions may typically be experienced in isolation, with no impact on the co-occurrence of other emotions. Human emotions can be represented as a network, wherein nodes represent specific emotions and the connections between them encode how likely emotions are to co-occur or inhibit one another. Graph theory can then be used to characterize and analyze this emotional network [
We characterized the centrality of the different emotions in the network by their
Finally, we note that the frequency of occurrence of specific emotions and their centrality in the emotional ecosystem are relatively independent. For example,
In order to explore potential gender differences in the centrality of the different emotions in the network, we analyzed the data separately for men and women (see
The line colors between specific emotions represent the extent to which emotions tend to co-occur (blue hues) or inhibit each other (red hues). The numbers in the grey dots underneath specific emotions represents their frequency of occurrence in the sample. The right panel represents the percentage of times respondents reported experiencing any, positive, negative, or mixed emotions.
The line colors between specific emotions represent the extent to which emotions tend to co-occur (blue hues) or inhibit each other (red hues). The numbers in the grey dots underneath specific emotions represents their frequency of occurrence in the sample. The right panel represents the percentage of times respondents reported experiencing any, positive, negative, or mixed emotions.
We sought to capture and characterize people’s everyday emotional experiences through an experience sampling smartphone application. Our findings revealed that everyday human life is profoundly emotional: people reported experiencing at least one emotion 90% of the time. Positive emotions were reported over 2.5 times more frequently than negative emotions. This finding is consistent with previous studies that aimed to capture everyday emotional experience [
First, future research may draw from our frequency findings to determine which particular emotions deserve more research attention. Specifically, whereas some infrequently experienced emotions have received much research attention (e.g., fear is experienced 5% of the time but according to Google Scholar has been the focus of over 100,000 articles), some frequently experienced emotions might be relatively under-researched (e.g., sadness is experienced 20% of the time and has been the focus of less than 3,000 articles on Google Scholar).
Second, future research may use our centrality findings to determine which particular emotions could be used as leverage in psychological interventions. Specifically, our network analyses suggest that some positive emotions (e.g., joy and satisfaction) are likely to inhibit the occurrence of many negative emotions, whereas other positive emotions (e.g., hope and gratitude) do not appear to show these buffering properties. A growing number of interventions promote the cultivation of specific emotional states like hope [
Although our findings break new grounds in several ways, the present research also suffers from several limitations, which should be addressed in future studies. First, although our smartphone application was designed to capture the widest possible range of episodes of daily life, participants had the opportunity to skip questionnaires. It is therefore theoretically possible that several emotions might be under- or over-represented, as different specific emotions might have different effect on people’s motivation to respond to the questionnaire prompts. In addition, responses to the app prompts were not evenly distributed throughout the day: essentially, the workday was oversampled. Also, it is theoretically possible that the specific emotion respondents experienced affected their likelihood of responding to the prompt. A second limitation lies in the dichotomous format of our emotion items. We chose to present our 18 emotions as a non-exclusive choice list. This allowed us to collect a very large amount of data, since participants would have been unlikely to respond as often if they had to rate each of the 18 emotions on continuous scales several times a day. However, we cannot exclude the possibility that dichotomous items required respondents to make idiosyncratic judgments about when to report an emotion as being present, which might have increased demand characteristics and the frequency of emotion reporting. Finally, although our choice of emotions respondents were presented with was based on careful consideration of the literature, the selected emotions may still be argued to differ in the extent to which they contain emotional vs. cognitive and attributional components. Future research may investigate to what extent emotion frequency and centrality are related to the relative importance of emotional vs. cognitive and attributional components of specific emotions.
Although our aim in the current research was not to focus on the structure of affect, our data indeed suggest that people generally experience pleasant emotions in the presence of pleasant emotions and unpleasant emotions in the presence of unpleasant emotions. In that sense, our findings align with one of the most well-known models on the structure of affect—the circumplex model [
Regarding our finding that some emotions inhibit each other, we feel we should clarify what ‘inhibition’ means in our cross-sectional dataset. When two emotions are negatively correlated, this either means that (1) one emotion inhibits the other, or (2) another variable inhibits one and stimulates the other emotion. In both cases, inhibition occurs, but information about the direction is lacking.
Further work is also needed to explore the connections between our findings about the centrality of emotions—and in particular the fact that men’s emotions were more strongly interconnected than women’s emotions—and related concepts of emotion granularity and emodiversity. Emotion granularity refers to people’s tendency to
Although research on emotions is abundant, knowledge about emotions in everyday life has been particularly scarce. Providing both basic foundations and novel tools, these findings provide evidence that emotions are ubiquitous in everyday life and can exist both in concert and distinctly, which has important implications for emotional interventions and theory.
(DOCX)
The authors are grateful to Jennifer Jordan, Yating Le, Hans Risselada, Yves-Alexandre de Monjoye, and Martin Desseilles for their help during various stages of the research. JQ was supported by the Barcelona School of Management and grant PSI2013-41909-P from the Spanish Ministry of Science and Education. MT was supported by the Helaers Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.