Psychological studies on Urami

Systematic psychological research on urami has been conducted against the backdrop of interest in stalking-related behaviors. An interview study’s results defined urami as a psychological state in which a person (1) feels that there is nothing one can do about an unpleasant situation caused by the words of another person (helplessness), (2) believes that the words of the other person are injustice (sense of injustice), and (3) cannot forgive the assailant (unforgiveness) [1]. A “trait urami scale” with factors corresponding to these three components of helplessness, injustice, and unforgiveness was developed and validated [9]. Furthermore, the trait urami correlated with higher levels of rumination and difficulties in emotion regulation through reappraisal [10]. However, studies using this scale are still limited, and the examination of its components and scale development were conducted based on a small sample size of 14 [1]. Therefore, it is necessary to examine the generalizability of the conceptual structure through the present study.

In contrast, Suzuki [2] asked participants to freely describe the concept and context of urami, attempting to examine the conceptual structure and situations involving urami through text mining. Text mining for several hundred samples ensures a certain extent of generalizability and captures the detailed responses of participants [11]. However, no studies have used such several hundred text data to examine the relationship between urami experience and individual differences among participants, as well as closely related concepts such as anger. Thus, we applied structural topic modeling (STM) [12], a novel algorithm in natural language processing (NLP), to precisely understand the concept of urami and the situations involving it by quantifying text data and the relationships with other variables.

Individual traits that can characterize Urami experiences

In addition to the frequency of urami experiences [2,9], the situations in which individuals are particularly likely to experience urami are also important indicators for evaluating individual differences. Examining situations in which participants are likely to experience urami based on their individual traits may provide valuable insights into the risks of stalking and other forms of aggressive behavior. In this study, we focused on the following five individual traits that may be related to urami and examined the correlations between each scale score and the descriptions of situations involving urami.

The first trait was interoceptive awareness. Interoception is the sensations that arise from within one’s body [13,14]. The importance and perception of physical reactions in emotional experiences and emotion regulation have been previously described [15]. A hypothetical model reported that excessive perception of noise in physical reactions results in persistent negative emotions, such as depression and anxiety [16], and that individuals who can accurately perceive their own heartbeat can appropriately control their emotions when they experience social exclusion [17]. Based on the findings regarding the relationship between interoceptive awareness and negative emotions in social situations, it is believed that indicators such as the inaccuracy of interoception and excessive attention to interoception are related to the experience of urami.

The second trait was narcissism. Stalking perpetrators are often diagnosed with Group B personality disorders, which are characterized by emotional and behavioral instability, including narcissism [18,19]. Perpetrators’ narcissistic personality disorder tendencies are particularly strongly correlated with the desire to threaten victims [20]. Traditional theories of narcissism have emphasized its oblivious aspect (characterized by a lack of concern for others [21]), however, in recent years narcissism has been understood as a two-dimensional concept that also includes a hypervigilant aspect (characterized by sensitivity to evaluation from others [22]). Although many unknowns are involved in the mechanisms behind this relationship, the greater the oblivious narcissistic tendency or the hypervigilant narcissistic tendency, the more likely it is that a person will feel urami toward the opponent, which may trigger stalking. A high level of narcissism is particularly likely to be related to urami in romantic relationships.

The third, fourth, and fifth traits were motivation for revenge, forgiveness, and self-efficacy, respectively. Motivation for revenge is the drive to eliminate the person who has caused one harm [23,24]. Forgiveness is treated as the opposite of urami, resentment, and grudges [25,26]. Self-efficacy refers to “people’s judgments of their capabilities to organize and execute courses of action required to attain designated types of performances” [27]. In particular, the term “generalized self-efficacy” refers to behavior in more generalized, everyday situations over the long term [24]. These three individual traits or scales were used to validate the trait urami scale [6]. These three concepts respectively correspond to the components of urami (helplessness, injustice, and unforgiveness) [1,9]. The scores of the trait urami scale positively correlated with the motivation for revenge and negatively correlate with forgiveness tendencies and self-efficacy [9].

STM

To conduct an effective analysis of text data, we used STM [12]. Topic modeling is a general term for NLP algorithms that identify “topics” that characterize documents. In latent Dirichlet allocation, which is the primary method used, documents are considered to be composed of multiple topics [28,29]. As topics are expressed as clusters of words that co-occur within a document, the proportion of each topic contained in that document can be calculated while simultaneously extracting each topic. A major feature of the STM, which is an advanced version of this, is the ability to incorporate metadata (external variables) linked to documents into the analysis as topical prevalence covariates, reflecting the ease with which topics appear [12]. In STM, it is possible to directly model the relationship between metadata and topics.

Although only a few studies have used STM in psychology and related fields, some of them have attempted to explore the nature of concepts based on text data obtained from participants. For example, relationships between a topic and rating scale scores regarding “awe” [30], thoughts about driverless vehicles, and individual traits such as the desire for independence and locus of control [31], as well as the distinguishability between generosity and fairness and the component of “virtue” [32], have been examined by STM. Several other psychological studies have also used STM to examine the relationship between writer attributes and topics [33,34].

When asking people to write freely about their emotions or thoughts, as done in this study, the content of their responses may be influenced by their verbalization skills. However, it is possible to examine this influence by considering each individual’s verbalization skills as a prevalent covariate. As it is possible to relate free-description data, which contain rich information, to quantitative variables, STM is an attractive method for researchers in the fields of emotion and cognition. By using STM, researchers can manage large amounts of descriptive data objectively and quantitatively.

Present study

This study examined the following: (1) the conceptual structure of and situations involving urami, (2) the relationship between urami situations and individual traits, and (3) the discriminability between urami and anger through analyses of text data. As this study is positioned as exploratory research to grasp the characteristics of urami, no specific hypothesis was set. The results were expected to correspond to the abovementioned aims using STM. First, based on the idea that urami is a higher-order emotion with multiple factors, we predicted that multiple interpretable topics reflecting these factors would be extracted from the data describing the concept of and situations involving urami. Second, we predicted that the proportion of the extracted topics of urami would correlate with each rating scale score. Third, we predicted that topics specific to urami would be extracted by comparing the described contents of urami and anger.

Participants

We conducted an online survey of 560 people using CrowdWorks, a crowdsourcing service (https://crowdworks.jp). People who speak Japanese as their native language and have experienced feelings of urami and anger were recruited to receive valid responses. Whether or not they met these criteria was self-reported. There were no other inclusion criteria. The sample size was determined for the extraction of topics using STM, the main analysis of this study. This study’s sample size was determined based on a previous study that obtained results with sufficient validity for a sample size of approximately 500 individuals [30]. The target sample size was 500, but there was a worker who used 50 accounts to receive rewards fraudulently, resulting in a non-negligible number of missing data. Additional recruitment was carried out to ensure the sample size was as originally planned. After excluding those who provided duplicate responses, did not pass the attention check, and provided free-description data obviously unrelated to urami or anger, the data of 489 participants (males: 247, females: 237; do not want to answer: 5; mean age: 42.32, standard deviation [SD]: 10.08; age range: 19–72) were analyzed. The participants were paid 400 yen as monetary compensation via CrowdWorks.

Open-ended items

Scales

Procedure

The survey was created and conducted using an online survey system (Qualtrics, https://www.qualtrics.com). Participants read the instructions regarding the survey, and only those who provided written informed consent answered all the questions. The recruitment and data collection for the survey were conducted from August 22 to August 29, 2024. At the beginning of the survey, we asked the participants to provide their age and gender. Participants could choose “male”, “female”, “other”, and “do not want to answer.” They then responded to free-description items pertaining to the concepts, situations, and physical sensations related to urami and anger. Additionally, they answered the seven abovementioned rating scales. The order in which the emotional conditions and scales were presented was randomized for each participant. In the Interoceptive Accuracy Scale (IAS) and the Interoceptive Attention Scale (IATS), an item stating, “For this item, please select (1) ‘strongly disagree’” was inserted as an attention check. The survey took approximately 25 minutes.

Data analyses

The main analysis was conducted using the statistical software R [47], following the approach of [30]. The RMeCab package [48] was used to preprocess the Japanese text data. The docDF() function was employed to extract parts of speech that carried interpretable meanings (verbs, nouns, adjectives, and adverbs), and the inflected words were unified into their root forms. Additionally, words that could become noise owing to their frequent appearance in various contexts (such as “する [do],” “なる [become],” “ない [not],” “いる [be],” and “ある [exist]”) were removed as stop words.

STM was performed using the “stm” package [49]. Consistent with prior recommendations [49], spectral initialization was used for model initialization. A default cutoff value of 500 was used to terminate the EM (Expectation–Maximization) iterations when the model did not converge. The procedure for determining the total number of topics (K) was explored by following the guidelines and using the searchK() function [50]. The range of K explored covered all integers between 3 and 20. For all these models, four metrics (semantic coherence, exclusivity, variational lower bound, and residuals) were plotted to evaluate the overall quality of the topics. The optimal number of topics for each analysis was determined based on these metrics. The topic-naming procedure also followed the guidelines of Weston et al. [50]. The first method for naming topics involved checking the words with high occurrence probabilities (β) for each topic, and the second method involved examining the content of documents with high topic proportions (θ) for each topic. For each topic, the words with high occurrence probabilities and documents with high topic proportions were output and the topics were named by reviewing these words and descriptions.

Two-stage STM analyses were performed on text data of concepts and situations. The first stage of STM addressed the first research aim and focused solely on the text data related to urami. In the STM analysis of the concept of urami, FFMQ-d was included as a prevalence covariate. In the STM analysis of the situations involving urami, in addition to age and the FFMQ-d, all scale scores were included as prevalence covariates to examine their relationships (as described later).

Next, regression analysis was conducted using the estimateEffect() function, with scale scores as independent variables and topic proportions obtained from the STM of the situations involving urami as dependent variables. As the focus was on the direct correlational relationships between each scale score and topic proportions, simple regression models were used to examine how each scale score predicted the topic proportions. In this analysis, the Benjamini-Hochberg procedure [51] was used with a false discovery rate (FDR) of 0.05 to account for the increased risk of Type I error due to multiple testing. This analysis addressed the second research aim.

In the second stage of the STM, urami and anger text data regarding concepts and situations were analyzed. The model included described emotions (urami or anger), age, FFMQ-d scores, and subjective distinguishability (degree to which urami and anger are subjectively distinguished). Age, FFMQ-d scores and subjective distinguishability were treated as control variables. The estimateEffect() function was used to examine the effects of these three variables. FDR correction [51] with the same conditions as above was applied.

Thus, this study examined four topic models corresponding to each research aim: the concept of urami (aim 1), situations involving urami (aims 1 and 2), the concept of urami and anger (aim 3), and situations involving urami and anger (aim 3).

Components of Urami

For the text data regarding the concept of urami, after reviewing the four indicators (semantic coherence, exclusivity, variational lower bound, and residuals), the optimal number of topics was determined to be nine. Plots of each indicator are depicted in S1 Fig-A in the appendix. The naming results for the nine topics identified by STM, as well as the words with high occurrence probabilities (β) for each topic, are presented in Fig 1-A and S2 Fig-A. The documents with high topic proportions (θ) for each topic are listed in S2 Table in the appendix.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Words with high occurrence probabilities for each topic.

The original Japanese word list was translated into English, and a word cloud was generated. As “emotion (感情)” was assigned to many topics, it was excluded from the word cloud to avoid hindering the interpretation of the topics. Panels A and B represent the STM results for the concept of and situations involving urami, respectively. The closer the hue is to blue, the higher the occurrence probabilities.

https://doi.org/10.1371/journal.pone.0349193.g001

The FDR correction was applied to 18 tests based on the combination of nine topics and two control variables (i.e., age and FFMQ-d score). Age positively predicted the proportion of Topic 3 (unreasonable / injustice; b = 0.002, SE = 0.00, t = 2.85, p = .030), and negatively predicted the proportion of 4 (negative emotion; b = −0.002, SE = 0.00, t = −2.66, p = .036) and 5 (persistence; b = −0.003, SE = 0.00, t = −3.61, p < .05). These results indicated that Topic 3 (unreasonable / injustice) was more likely to be discussed by older participants, whereas Topic 4 (negative emotion) and 5 (persistence) were more likely to be discussed by younger participants. The FFMQ-d score negatively predicted the proportion of Topic 3 (negative emotion; b = −0.03, SE = 0.01, t = −2.81, p = .030). The effects of these variables in other topics did not survive the FDR correction (ps > .050).

Situations involving Urami

For the text data regarding situations involving urami, based on the indicators (S1 Fig-B in the appendix), six topics were determined to be the optimal number. Fig 1-B and S2 Fig-B presents the naming results for each topic identified by STM, along with words that have high occurrence probabilities (β) for each topic. S3 Table in the appendix lists the documents with high topic proportions (θ) for each topic. The effects of control variables were mentioned in the next section.

Relationships with rating scale scores

Table 1 presents the descriptive statistics for each scale score. The results of the regression analysis, with topic proportions related to the situations involving urami as the dependent variable, are presented in S4 Table in the appendix. The FDR correction was applied to 54 tests based on the combination of six topics and nine covariates. Regarding control variables, age positively predicted the proportion of Topic 3 (harm to family / troubles with money) and negatively predicted the proportion of Topic 4 (being bullied at school / cold treatments). The FFMQ-d score positively predicted the proportion of Topic 3 (harm to family / troubles with money).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Mean Score and Standard Deviation (SD) of each rating scale.

https://doi.org/10.1371/journal.pone.0349193.t001

Regarding target variables, interoceptive attention positively predicted the proportion of Topic 6 (violence / back biting). Hypervigilant narcissism positively predicted the proportion of Topic 6 (violence / back biting) and negatively predicted the proportion of Topic 3 (harm to family / troubles with money). Motivation for revenge negatively predicted the proportion of Topic 3 (harm to family / troubles with money). Self-efficacy negatively predicted the proportion of Topic 6 (violence / back biting).The effects of other variables did not survive the FDR correction (corrected ps > .050).

Distinctiveness in components

For the text data on the concepts of urami and anger, the optimal number of topics was determined to be 11 (see S3 Fig-A in the appendix for each indicator). The FDR correction was applied to 44 tests based on the combination of 11 topics and four covariates (i.e., emotional condition, age, FFMQ-d score, and subjective distinguishability).

Based on the descriptions with high topic proportions θ identified by STM (presented in S5 Table in the appendix), Fig 2 shows the names of the topics and their corresponding emotional condition effects. S4 Fig-A and S5 Fig-A in the appendix shows the words with the highest occurrence probabilities for each topic. In the texts describing urami, Topic 2 (desiring assailant’s unhappy / intention to revenge; b = 0.12, SE = 0.01, t = 13.24, p < .001), 5 (persistence / unforgiveness; b = 0.18, SE = 0.01, t = 17.86, p < .001), 7 (intense negative emotion; b = 0.08, SE = 0.01, t = 9.32, p < .001), and 9 (disgusting / disappointment; b = 0.02, SE = 0.01, t = 2.48, p = .041) appeared more frequently. In contrast, in the texts describing anger, Topic 1 (injustice / unreasonable; b = −0.07, SE = 0.01, t = −8.95, p < .001), 3 (arising feeling; b = −0.06, SE = 0.01, t = −8.24, p < .001), 4 (being irritated; b = −0.06, SE = 0.01, t = −6.58, p < .001), 8 (helplessness / having nowhere to direct one’s emotions; b = −0.04, SE = 0.01, t = −4.79, p < .001), 10 (irritation / arousal / bodily sensations; b = −0.13, SE = 0.01, t = −13.35, p < .001), 11 (suddenness / uncontrollable; b = −0.06, SE = 0.01, t = −6.49, p < .001) appeared more frequently.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Naming results for each topic related to the concepts of urami and anger, and the effects of emotional conditions.

The points plotted for each topic represent the regression coefficients and the error bars depict the 95% confidence intervals. Topics that significantly appeared more frequently when participants were asked to describe urami and anger are marked in red and blue, respectively. The light blue and red colors indicate that the test results did not survive the FDR correction.

https://doi.org/10.1371/journal.pone.0349193.g002

Regarding control variables, age positively predicted the proportion of Topic 6 (sense of being harmed by betrayal or a violation of one’s dignity; b = 0.002, SE = 0.001, t = 2.89, p = .015), and negatively predicted the proportion of Topic 4 (being irritated; b = −0.001, SE = 0.0004, t = −3.14, p = .008) and 9 (disgusting / disappointment; b = −0.001, SE = 0.0003, t = −2.65, p = .027). The FFMQ-d score negatively predicted the proportion of Topic 8 (helplessness / having nowhere to direct one’s emotions; b = −0.02, SE = 0.004, t = −4.39, p < .001). The effects of other covariates including subjective distinguishability did not survive the FDR correction (ps > .050).

Distinctiveness in situations

For the text data on the situations involving urami and anger, the optimal number of topics was determined to be 11 (see S3 Fig-B in the appendix for each indicator). The FDR correction was applied to 44 tests based on the combination of 11 topics and four covariates.

Considering the topics identified through STM, topic names were determined based on the content with the highest topic proportions (S6 Table in the appendix). Fig 3 illustrates the influence of emotional conditions on each topic. Additionally, S4 Fig-B and S5 Fig-B in the appendix shows the words with the highest occurrence probabilities for each topic. In the texts describing urami, Topic 2 (getting bullied; b = 0.10, SE = 0.02, t = 6.58, p < .001) appeared more frequently. In contrast, in the texts describing anger, Topic 3 (sighting of inappropriate behaviors; b = −0.12, SE = 0.01, t = −9.03, p < .001) appeared more frequently.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Naming results for each topic related to the occurrence situations of urami and anger, and the effects of emotional conditions.

The points plotted for each topic represent the regression coefficients and the error bars depict the 95% confidence intervals. Topics that significantly appeared more frequently when participants were asked to describe urami and anger are marked in red and blue, respectively. The light blue and red colors indicate that the test results did not survive the FDR correction.

https://doi.org/10.1371/journal.pone.0349193.g003

Regarding control variables, age negatively predicted the proportion of Topic 2 (; b = −0.004, SE = 0.001, t = −5.30, p < .001). The effects of other covariates did not survive the FDR correction (ps > .050).

The results of the paired t-test indicate that the time when the event that triggered the feeling of urami occurred (mean 13.37 years ago) was significantly earlier than the time when anger occurred (mean 8.58 years ago; t(488) = 9.01, p < .001, d = 0.41).

Components of and situations involving Urami

The topic modeling extracted meaningful topics that are consistent with previous study that identified “injustice”, “unforgiveness” and “helplessness” as components of urami [1,6]. Among the topics identified from the descriptions regarding the concept of urami, “unreasonable / injustice” corresponds to “injustice,” while “desiring assailant’s unhappy” and “unforgiveness / intention to revenge” correspond to “unforgiveness.” In this study, the topics corresponding to “unforgiveness” were subdivided, and two new topics were identified. Topics named “persistence / helplessness” is thought to correspond to “helplessness”. Interestingly, the number of topics named “persistence” was emphasized. “Unforgiveness”, although labeled differently, may also include such elements of persistence. From the perspective of the persistence of emotion [52], two distinct patterns emerged: one involving the motivation for revenge against the perpetrator and another focused solely on desiring the perpetrator’s unhappiness. The newly identified “negative affect” reflects emotional valence, and “sense of being harmed” reflects the condition of occurrence, both of which can be interpreted as topics representing fundamental elements that constitute emotion. Additionally, these topics demonstrate consistency with previous definitions and indicate the existence of multiple subtypes of urami that differ in their other-oriented nature. “Cognitive load” reflects those who mentioning that the persistence of urami strains cognitive resources, suggesting that urami may reduce well-being. Unresolved events tend to come to mind even during unrelated tasks [53], and this topic is considered to reflect such a phenomenon.

Interpretation for the effects of control (non-target) variables are unpredicted and tentative. However, older participants tended to report Topic 3 (unreasonable / injustice) as concept of urami, whereas younger participants reported Topics 4 (negative emotion) and 5 (persistence). This may suggest that older participants are more likely to refer to higher-order concepts. Lower verbalization skills (FFMQ-d scores) were related to the prevalence of Topic 3, which may reflect that perceived injustice is a central component of urami that can be recognized even by those with lower verbalization skill.

The topics pertaining to situations involving urami indicate that urami arises in various settings, such as romantic relationships, workplaces, and schools. This aligns with findings from a prior study [2] that examined the occurrence of urami using similar procedures. Bullying experienced in these situations has been shown to cause significant damage to quality of life [54], and the persistent negative emotional state it triggers may be labeled urami. Furthermore, events that triggered urami tended to have occurred further in the past than those that triggered anger. Although these events were less frequent, their greater impact on life may have made them more vivid.

Relationships with individual traits

The analysis of the relationship between psychological scale scores and situations involving urami indicates that the likelihood of experiencing urami varies depending on certain individual traits. Participants with low levels of motivation for revenge tended to experience urami in the situations described in Topic 3 (harm to family / troubles with money). Topic 3 was also associated with lower levels of hypervigilant narcissism. In cases where the violation of one’s own dignity is minor—such as harm directed at a family member—people may be (relatively) less likely to feel a motivation to revenge.

Individuals with higher hypervigilant narcissism and lower self-efficacy were more likely to experience urami related to Topic 6 (violence backbiting). Individuals with high levels of hypervigilant narcissism, who are more sensitive to others’ evaluation [22,40], may be more prone to feel unpleasant when their self-image is threatened by violence and backbiting from others. This can lead to a stronger association with urami. It is interesting that individuals who pay greater attention to their internal bodily states (i.e., interoceptive signals) were more likely to feel urami in response to violence.

Regarding control variables, older participants showed higher prevalence of Topic 3 (harm to family / troubles with money), whereas younger participants showed higher prevalence of Topic 4 (being bullied at school / cold treatments). If individuals had multiple experiences of urami, this may be because more recent events were more likely to be reported. The association between verbalization skills and Topic 3 is difficult to interpret. However, as noted above, individuals with higher verbalization skills may be able to appropriately describe situations even when the violation of their own dignity is relatively minor.

The lack of significant effects for most variables may be due to the use of topic proportions as the dependent variable, rather than the intensity or frequency of emotional experiences. To the best of our knowledge, there are no studies that have examined the relationship between emotional triggers and individual traits using similar approach, which makes these interpretations tentative. Future research should experimentally manipulate situations and measure the intensity of urami elicited in each context to further investigate the relationship between urami experiences and individual differences.

Differences from anger

The analysis including emotional condition as a prevalence covariate showed that urami and anger differ in their components. When participants were asked to describe the concept of urami, topics related to persistence—such as “desiring assailant’s unhappy / intention to revenge”—were discussed more frequently. In contrast, when describing anger, topics related to “irritation / arousal / bodily sensations” or “suddenness / uncontrollable” were more prevalent. This strongly supports the claim that urami is an emotion characterized by persistence—a crucial factor in distinguishing and defining emotions [52,55]. A study that examined the duration of 27 different emotions reported that anger lasts for approximately 24 hours on average [55]. If a similar procedure were used to assess the duration of urami, it would likely be reported as a longer-lasting emotion, similar to hatred or sadness. Among the components of urami [9], injustice and helplessness appeared more frequently in anger. For these to constitute components of urami, they may need to be experienced persistently.

Regarding the control variables, as with the concepts of urami, younger participants tended to report relatively simple concepts such as Topics 4 (being irritated) and 9 (disgusting / disappointment), whereas older participants tended to report more complex concepts such as Topic 6 (sense of being harmed by betrayal or a violation of one’s dignity). Lower verbalization skills were associated with Topic 8 (helplessness/ having nowhere to direct one’s emotions), which may be because the content of Topic 8 reflects a central component of resentment that can be reported even by individuals with lower verbal ability. However, note that this interpretation is also tentative.

Unlike conceptual differences, only a few topics differed in prevalence between urami and anger in the situation. When participants were asked to describe situations involving urami, Topic 2 (getting bullied) was discussed more frequently. In contrast, when describing anger, Topic 3 (sighting of inappropriate behaviors) was discussed more frequently. Overall, urami tended to occur in situations where individuals were harmed by people with whom they had close or frequent interactions, such as classmates, and colleagues. In contrast, anger was more likely to arise in temporary relationships with less personal significance. When a perpetrator is close to someone, severing ties is often difficult. Continuing a relationship with the perpetrator after the harm has occurred may contribute to the persistence of negative emotions, ultimately leading to urami. Among the situational factors that influence emotion elicitation [56], anger has been suggested to arise when the perceived ability to change a situation is relatively low [57]. Urami may be associated with an even greater perceived helplessness compared to anger.

In both conceptual and situational analyses, some topics appeared with a similar frequency, regardless of whether the described emotion was urami or anger. This reflects a conceptual overlap between the two emotions, supporting the idea that urami is a higher-order emotion containing both unique components and components shared with anger.

Taking advantage of the strengths of STM [12], this study incorporated several control variables and quantified the magnitude of their effects. It was assumed that individual description skills may influence free descriptions; however, contrary to this assumption, their effect was not significant for all topics. Although we only recruited participants who had experienced urami toward others, it is possible that urami is a psychological state that is relatively easy to verbalize. Similarly, the effect of subjective distinguishability was not observed in any topics, suggesting that, even if individuals do not consciously differentiate between the terms urami and anger, they may still internally represent them as distinct emotional concepts. However, subjective distinguishability was measured with a single-item scale that is general and providing limited information; therefore, there might be a limitation to its interpretation.

Social implications

This study can be positioned as an initial approach toward addressing social issues related to urami. Urami may underlie various aggressive behaviors, such as stalking and homicide, and is likely rooted in the distress of persistent unforgiveness and perceived unfairness. While individual factors such as narcissism and self-efficacy are related, situational factors such as being bullied at school are also important and cannot be overlooked. Demonstrating this quantitatively using STM is a key contribution of this study, and it may provide evidence for social interventions aimed at preventing urami, such as education.

Limitations and future prospects

A key contribution of this study is its multifaceted approach to the concept of urami. However, some limitations must be considered. First, the naming of topics is inherently constrained by methodological factors, as it relies on researchers interpreting and labeling each topic based on individual descriptions.

Second, because this study analyzed free-description data, it is possible that only participants who were highly skilled at describing their thoughts were included. Although the self-reported describing skill measured by the FFMQ-d was only weakly associated with topic frequency and showed no extreme deviations, it remains to be seen whether similar results would be obtained using objective linguistic measures, such as vocabulary size. Regardless, developing a new psychological scale to assess the traits or states of urami based on collected text data could help explore how it manifests as a psychological state among individuals with lower verbalization skills.

Third, the number of concepts/variables to examine relationships with urami was limited. The association with narcissism highlights the social-emotional aspect of urami. It would be interesting to extend this finding by examining its relationships with social psychological process such as empathy or self-consciousness.

Finally, although urami and anger were suggested to be fundamentally distinguished by their persistence, this study focused more on the conceptualization of emotions rather than on their temporal dynamics. As the next step, the chronological relationship between urami and anger should be examined. Specifically, negative emotions arising in similar situations may initially manifest as anger and later be recognized as urami after a certain period. This temporal shift may explain the observed overlap in descriptive data across conditions.