Linguistic Affect: Positive and Negative Emotion Words are Contagious, Predict Likability, and Moderate Positive and Negative Affect
IN THIS ARTICLE
Positive affect (PA) is active, enthusiastic, and happy engagement in pleasurable activities and negative affect (NA) includes aversiveness, anger, and fear (Watson et al., 1988). Two studies examined linguistic affect presented as emotion words used to describe experiences with PA and NA. The first study explored linguistic affect priming and altruistic decision-making, PA and NA valence word-choice, likability, and affect. 132 undergraduates were randomly assigned to read a narrative with positive or negative linguistic affect priming. Altruism was assessed and no difference in altruistic decisions existed between linguistic affect conditions, however, participants exposed to positive linguistic affect used more PA valence words (p = .005*) and those in the negative linguistic affect condition used more NA valence words (p = .002*). Responses to Reysen’s Likability Scale (Reysen, 2005) indicated the narrative written from a positive perspective produced an individual that was more likable (p ≤ .001*) than the narrative written from the negative perspective. The Positive and Negative Affect Scale (Watson et al., 1988) showed no difference in affect scores between linguistic affect conditions. A second study of 108 undergraduates examined whether affect was better predicted by (a) the PA or NA experience written or (b) the positive or negative emotion words provided for use. A two-way ANOVA showed a main effect (p = .002*) for the valence of words used among the four conditions. Linguistic affect exposure impacts word-choice and likability and using positive or negative emotion words to describe experiences regulates the affect states incurred from writing about personal experiences.
Language is often used to communicate emotional reactions from events to assist in obtaining our physical or emotional desires. Humans have adaptively developed emotions to direct our attention to stimuli in the environment which could impact our well-being (Ekman, 2003). Barrett (2017) argues our brains construct our experience of emotion through predictive interactions and that words help babies create constructs to understand the physical world. It was only recently that researchers began to study the impact of positive and negative linguistics on emotional mental health and “negative” research still outnumbers “positive” (Mayne, 1999). The field of psychology is relatively new and the idea of positive psychology is even newer still. According to Seligman and Csikszentmihalyi (2000), positive individual traits allow people to maintain a higher quality of life. Too much of the psychological focus, in their view, has been focused on negative information, pathologies, diagnoses, and victimology, which has left the field of psychology out of balance and out of touch with human strengths and virtues, impacting society at large (Seligman & Csikszentmihalyi, 2000). Seligman discusses a human tendency to focus on negative information as an evolutionary adaptation from the past (Seligman, 2002; Seligman, 2011). That adaptation may no longer serve us in a world where we experience constant stressors from modern life.
Many individuals within the field of therapy have learned to focus on pathology and diagnoses (Freedman & Combs, 1996). The medical model has focused on disease and the educational system has created helpers who are often overly focused on being experts (Freedman & Combs, 1996; Small, 1989). By focusing on negative information and pathologies, the field of psychology had lost focus on creating fulfilled individuals and facilitating thriving societies (Seligman & Csikszentmihalyi, 2000). Nurturing strengths through positivity training rather than punishing mistakes with negative language and pathology labeling may foster traits of courageousness, optimism, hope, honesty, perseverance, work ethic, insight, and creative flow capabilities which can both defend against and help individuals overcome mental illness (Seligman & Csikszentmihalyi, 2000).
Linguistics in Psychology
Cognitive behavioral therapy (CBT) is a commonly used therapeutic technique for correcting maladaptive or negative thought patterns within individuals by actively changing negative thought patterns into more positive self-perceptions (Brannon et al., 2017). CBT aims to develop beliefs, thoughts, attitudes, and skills which can help create positive changes in behavior (Brannon et al., 2017). This psychological technique focuses on positive mental health outcomes by promoting positivity and optimism in clients. Positive and optimistic individuals tend to maintain increased qualities of life (Seligman & Csikszenmihalyi, 2000).
Narrative therapy attempts to reform the meaning behind life narratives in a more positive light to help clients (Freedman & Combs, 1996). Narrative therapists use remembering interviews for corrective experiences to change negative interpretations which bring about distressing emotional reactions to prior events (Freedman & Combs, 1996). This type of therapy takes a relational worldview in which each of our identities are fluid. According to this relational view, our identities are the actions we take; it is a process, not a possession (Freedman & Combs, 1996). The relational worldview provides support that we have a choice over the language we choose to use and how we interpret our environment. We have a choice over the words we use to identify others, ourselves, or any other experiences that may elicit emotional stress in order to actively construct positive emotions and identities.
The common and vital theme among these aspects of mental health tools is language use by clients. Positive or negative language is an important part of life and can shape our self-perceptions and evaluations of others (Lineweaver & Brolsma, 2014; Martijn et al., 1992). Lewine et al. (2018) found evidence that women from harsh economic backgrounds greatly benefited from positive affect (PA) measured in their use of language.
Tauszic and Pennebaker (2018) developed the Linguistic Inquiry Word Count (LIWC) program to evaluate linguistic responses, with evidence that emotional words provide important cues to thought processes and emotional states. These researchers explored the idea that writing about emotional events is beneficial (Tauszic & Pennebaker, 2018). Other studies have found language is an important tool in mental health, especially when discussing PA and negative affect (NA).
Affect can be described as the general sense of how an individual feels throughout the day and is a core aspect of our consciousness (Barrett, 2017). The study of affect and linguistics may be a way to understand how we construct our emotions. Feelings of pleasantness or unpleasantness are described as valence of affect (Barrett, 2017). PA is related to happiness and extraversion levels (Bobić et al., 2015), while negativity is linked with instability (Dizén & Berenbaum, 2011) and can be a determinant of unexplained medical complaints (De Gucht et al., 2004).
PA and NA are made of many distinct dimensions of emotion (Watson et al., 1988), but affect is not the same as emotion (Barrett, 2017). PA is defined by the extent to which an individual is active, alert, enthusiastic, happy, engages in pleasurable activities, and has been linked to higher engagement with social activities (Watson et al., 1988). NA is a dimension of distress and includes mood states of hostility, aversiveness, anger, contempt, guilt, fear, nervousness, and is associated with emotional distress including anxiety and depression (Watson et al., 1988).
NA has been linked with higher rates of emotional reactivity to stress and an increased vulnerability to disease (Mendonça-de-Souza et al., 2007) which supports prior research that negative emotional responses often reflect a state of illness while positive emotions correlate with health and well-being (Mayne, 1999). Research has indicated that personality differences may be linked to differences in PA and NA (Bobić et al., 2015).
Positivity-Offset and Negativity-Bias
Ito and Cacioppo (2005) described the positivity-offset as a motivational approach system which is activated in a separate part of the brain than the negativity-bias, a withdrawal motivational system. Ito and Cacioppo (2005) presented positive and negative pictures to participants, paired with both negative and positive language in the pictures. The results of this study indicated the existence of personal differences in positivity and negativity-biases (Ito & Cacioppo, 2005). This may be due to the dependency of biases on mood; whatever bias an individual is experiencing at that moment may lead them to recall either positive or negative memories which, then, reinforce their current mood (Lineweaver & Brolsma, 2014). This could be explained by the brains’ predictive capabilities (Barrett, 2017). Affective realism is the idea that our PA or NA can influence judgments about the world around us, meaning that current affect can influence people to believe objects or individuals in their environments are essentially positive or negative (Barrett, 2017).
Lineweaver and Brolsma (2014) created a memory self-efficacy survey with questions framed positively, negatively, or neutrally and compared self-efficacy with current mood states of participants. Positive moods were correlated with increased positive self-perception (Lineweaver & Brolsma, 2014). Depression and social desirability also mediated between affect and memory self-efficacy (Lineweaver & Brolsma, 2014). Interestingly, participants were more likely to rate themselves harshly in positively worded questions, which could be because boasting in an already positive environment may be viewed as socially undesirable (Lineweaver & Brolsma, 2014; Wenhui et al., 2016).
Baolin et al. (2010) compared arousal of emotional reactivity by priming with words and pictures. These researchers measured brain activity using electroencephalogram (EEG) neurocognitive scans to provide evidence that arousal activation from picture only and word only conditions were comparable (Baolin et al., 2010). Could the words we use and are exposed to everyday be as powerful as viewing images? It is an important side note that the way we interpret language is vital to understanding this process (Freedman & Combs, 1996). Indeed different cultures have different ways of describing emotions, with some having multiple levels of anger or identifying anger not as an individual emotion, but an interaction between two parties (Barrett, 2017).
Decision tasks were measured by Matthews et al. (1995) in groups presented with positive or negative words. The negative word group showed faster reaction times during decision making tasks (Matthews et al, 1995). Matthews et al. (1995) attributed this to the fact that humans are hard-wired to detect negative information with priority as it may be connected to a potential hazard or danger. This supports other research which indicates the negativity-bias is an evolutionary adaptation to avoid danger (Ito & Cacioppo, 2005; Seligman, 2006).
Our ability to quickly detect negative information suggests negative information pathways may be more readily available for some (Mathews et al., 1995). This supports the evolutionary explanation for the negativity-bias and why it remains so prevalent in modern times even though society also shows a need for more positive thinking and focus on the virtues of humankind (Seligman & Csikszentmihalyi, 2000). Animals and humans both react and learn more quickly about negative events, which have been shown to be more psychologically potent (memorable) than positive events (Rozin & Royzman, 2001). Though humans remember negative events quickly, Rozin and Royzman (2001) also asserted that, across cultures coming from 20 different languages, humans have more positive adjectives and positive descriptive words than negative ones. This could indicate humans desire states of positive experience and PA more so than experiences that elicit distress and NA. Despite this, negative information often elicits a greater response than positive information. Mendonça-de-Souza et al. (2007) measured glutticorticoids present in saliva prior to and following exposure to negative or unpleasant photos and showed exposure to unpleasant photos increased the hormonal stress response following the stressful activity. This could explain why negative information has been shown to be more influential in the formation of evaluative personality judgments than positive information (Martijn et al., 1992). Stress hormones impact these evaluations and memories during states of NA.
Birmingham et al. (2009) explored positivity and negativity-biases by examining the effects of experimenter language on cardiovascular stress responses. The four independent groups were supportive (high-positive), aversive (high-negative), ambivalent (high-positive, high-negative), and indifferent (low-positive, low-negative; Birmingham et al., 2009). The highly positive experimenter was found to be more appreciative, friendly, and helpful, whereas the highly negative experimenter was found to be more upsetting and dominant by participants, especially the women (Birmingham et al., 2009).
Birmingham et al. (2009) provided evidence that the ambivalent group experienced the most cardiovascular stress reactivity. This is likely credited to the fact that the more random and unexpected events are, the less perceived control, and greater amount of stress the event may cause. Other researchers provided evidence that ambiguity has a positive correlation with negativity-bias activation, while low ambiguity was associated with a positivity-offset (Gibbons et al., 2016). This confirms research by Ito and Cacioppo (2005) who found the positive motivational system was more likely to respond when evaluative input was low, and the negativity-bias responds more intensely during increased evaluative information. This begs the question of whether or not lowering emotional ambiguity by mindfully choosing positive language to describe events and, therefore, reducing negativity-bias, might increase operation of our positivity-offset.
Language Training Modifies Cognitive Biases
Becker et al. (2015) studied positive and negative language use through cognitive bias modification which aims to alter thought processes no longer conducive to an individual’s life. Positive and negative pictures were shown to 141 participants (Becker et al., 2015). The positive group represented positivity training with those viewing positive pictures experiencing less stress (Becker et al., 2015). Interestingly, the results also showed negative training could reverse a positivity-offset, which again, highlights the importance of negative or positive information exposure (Becker et al., 2015). The positive training group showed an increased positivity-offset and the participants primed with positive words were more likely to select positive pictures, which implies that the mere presence of information can impact individual decision making (Becker et al., 2015).
Crum (2016) reported that dialectic priming with rational beliefs could reverse irrational belief systems, showing that the tool of positive priming can reverse negative thinking patterns and vice-versa. The studies highlighted here (Becker et al., 2015; Crum, 2016) indicate how linguistics can influence positivity-offsets or negativity-biases and supports the need to back up these claims with more research driven evidence.
In healthy adults, the positivity-offset is seen as a way of being (Becker et al., 2015) and cognitive bias modification can help heal maladaptive thought patterns in individuals who are not naturally experiencing the positivity-offset. This is the long term goal of many therapeutic techniques. Depression is thought to be caused by too much negativity-bias activation and increased stress perception (Dizén & Berenbaum, 2011). Pleasant affect is associated with increased self-esteem and increased positive perception of how one is being treated by others (Dizén & Berenbaum, 2011). Maintaining PA can be difficult because affective states are often influenced by outside events which are out of our control. The language we choose to use however, is directly in our control and may provide a gateway to our positivity-offset or negativity-bias activation.
Some research on emotion indicates individuals may feel an emotion simply by remembering an event or even just by making certain facial expressions to bring about that emotion (Ekman, 2003). Is it possible that simply looking at a positive or negative word or using that word to describe events facilitates a re-experience of the PA or NA represented by the language used? According to Barrett (2017), this may be possible through the way we construct emotions, but it is only part of emotional development. Simply reading the word apple can elicit the brain to behave as if an apple is actually in your presence through simulation (Barrett, 2017). Linguistics may be one aspect of how we construct emotion by simulating past experiences of the words we are using.
Ekman (2003) discusses a complex affective system in the brain that helps bring about emotions yet is not completely understood. Herbert et al. (2013) found evidence that mindfulness in language use indeed plays a role in emotional processing when participants labeled their own emotions and others as angry, happy, or fearful indicating the importance of labelling. Herbert et al. (2013) recommended that the way individuals use language be studied to understand more about emotional processing.
Prior research (Becker et al., 2015; Crum, 2016) indicates the language we use reflects emotional reactivity and that words may be just as influential as images in eliciting brain activation (Baolin et al., 2010), however, there is a paucity of research specifically addressing the use of emotionally charged descriptive language to recount experiences, circumstances, or individuals which bring about feelings of PA and NA. This type of language is used commonly to describe life events. The term linguistic affect was coined to facilitate the study of the relationship between the way we use emotion words to describe experiences and the affective state we incur from those descriptions. Linguistic affect is operationally defined as emotion words used to describe PA and NA experiences. For the purposes of the current research, only positive and negative emotion words as associated with direct meaning was used (e.g., not sarcasm). Herein, two novel studies explored the idea of positive and negative linguistic affect. The first study exposed individuals to positive or negative linguistic affect priming through the reading of a pre-written narrative to determine the relationship with altruistic decision, PA or NA valence word-choice, likability, and affect. In a second study, participants wrote their own narrative using either positive or negative emotion words in retellings of PA or NA experiences to identify whether the affective experience being described, or the emotion words used to describe said events regulated affect states.
Study 1 – Exposure to Linguistic Affect Through Narrative Reading
Altruism (prosocial helping behavior) is an important part of psychology and has been linked with increased happiness and PA (Seligman, 2002). Birmingham et al. (2009) showed positive language increased ratings of friendliness and helpfulness of an experimenter giving direction. Prosocial behavior can increase belonging through increasing operation of the positivity-offset and prosocial behavior has also been linked to maintaining a positive outlook (Luengo Kanacri et al., 2017). Prosocial spending and happiness have also been found to run in a positive feedback loop (Aknin et al., 2012). Therefore, altruism was included in the first study.
Language usage in social interactions could impact those exposed to it and individuals may change their attitudes or behaviors if provided evidence of its measurable detrimental impacts to themselves or other individuals they are exposing to it. Specifically, Becker et al. (2009) found evidence that positivity training modified emotional vulnerability to NA and that positive or negative information exposure can reverse a positivity or negativity-bias. PA and NA may cause differences in evaluations of others (Machin & Jeffries, 2017; Martijn et al. 1992). Previous research has indicated humans have a need to belong and that likable individuals are highly sought after (Machin & Jeffries, 2017). For these reasons, PA and NA valence word-choice, likability, and affect were also measured.
There is extensive evidence for the benefits of positive thinking, positive language use, and the harmful effects of their negative counterparts. Previous research considered the impact of language on mental health and these results showed an apparent link between PA, NA, the associated biases, and negative word/picture priming (Boalin et al., 2010), cardiovascular stress-reactivity (Birmingham et al., 2009), reversal of biases with dialectical training (Crum, 2016), hormonal stress responses (Mendonça-de-Souza et al., 2007), and perception of health (Goodwin & Engstrom, 2002).
To our knowledge, a paucity of research exists examining whether presentation of narrative content through reading and processing of positive or negative linguistic affect can impact individuals in a variety of ways. The current study included positive or negative linguistic affect words in a conversational narrative to prime participants. This explored the impact of linguistic affect priming on four dependent variables: altruistic decision, word-choice within that decision making process, likability, and affect. It should be noted that Seligman (2006) has studied evaluative processes. The current study did not examine optimism or pessimism in this way, rather we examined the direct use of positive and negative emotion words to describe experiences of PA or NA.
● Ha1: Those exposed to positive linguistic affect will be more willing to help than those exposed to negative linguistic affect.
● Ha2: Those exposed to positive linguistic affect will be more likely to use PA valence words in their explanations of altruistic decisions, while those exposed to negative linguistic affect would use more NA valence words.
● Ha3: An individual shown to use positive linguistic affect will be rated more likable than the individual using negative linguistic affect.
● Ha4: Positive linguistic affect priming will increase PA and negative linguistic affect priming will increase NA.
Method & Survey Measures
Positive and negative linguistic affect narratives were created from the perspective of a fictional character named ‘Tomas.’ Participants were instructed this individual was a friend of theirs and that they had agreed to listen to him vent about an experience. The same experience was recounted in both conditions, but Tomas’ affect in the conversational narrative impacted his evaluation of the events and this was portrayed through the use of positive or negative emotion words when he explained his day in the conversational narrative. The independent variable, the positive or negative emotion words present in Tomas’ narrative, was defined as linguistic affect priming and the narrative was 442 words long.
This first study assessed participant altruism through response to a 2AFC question about offering help to Tomas, followed by an open-ended explanation of their decision. Answers to these explanations were evaluated using LIWC software (Tausczik & Pennebaker, 2009). These linguistic analyses evaluate language including word count and the measuring of many other variables, however, for our purposes, outputs of PA and NA emotional valence words in their explanation were the focus. LIWC outputs may be used to extrapolate to other variables such as emotional states and social relationships (Tausczik & Pennebaker, 2009).
Reysen’s Likability Scale (RLS; Reysen, 2005) was employed to examine likability as a function of linguistic affect. The RLS obtained evidence for convergent and divergent validity with levels ranging from .90 to .91 by using laughter and personality measurements to predict an accurate scale of likability in individuals (Reysen, 2005).
The Positive and Negative Affect Scale (PANAS; Watson et al., 1988) was used to measure participant affect and has been shown to be valid compared to other measures of state-anxiety, depression, perceived stress, overall health well-being, and has been shown reliable across college and adult populations. NA ratings in the PANAS correlated with both the Hopkins Symptom checklist of depression and Beck Depression Inventory (Watson et al., 1988). Perceived stress and NA ratings were also linked, but not to PA. PA was however linked with higher social activity whereas NA was not (Watson et al., 1988). The PANAS has been used to measure individual PA and NA across a variety of time frames and populations (Watson et al., 1988). These PA and NA ratings, therefore, may reflect other variables which have been shown to have a direct impact on psychological and physical health outcomes, for example distress (Brannon et al., 2017; Watson & Pennebaker; 1989). For materials see Appendix A.
The present research was conducted at a small western liberal arts university. Participants were 140 undergraduate students selected via convenience sampling who were tested in small classrooms and conference rooms. Although informed consent was collected, the purpose of this study was not disclosed prior to testing as knowing the purpose of the research could impact individual positivity or negativity-biases. Participants were informed all data collected was in no way connected to their identity, ensuring confidentiality of all participants and were told what would be expected from their participation. They were also advised of their ability to withdraw from the study at any time without penalty. Research credit was provided as compensation for participation.
A double-blind study design was used and random assignment was achieved via Microsoft Excel’s random number generator function. Test packets were assembled and placed in this random order prior to data collection and distributed as such so the experimenter was unaware of condition assignment. Participants completed test packets individually, but were tested in groups.
Participants were provided ten minutes to read their assigned narrative and complete questionnaires. Pre-testing indicated ten minutes was ample time for college undergraduates to complete the packet. The experiment was facilitated without impacting participant’s positivity-offset or negativity-bias by remaining neutral during data collection sessions, as previous research does show positive or negative speaking tone can impact participant responses (Birmingham et al., 2009).
Participants in two conditions were primed by reading narratives from the perspective of ‘your friend Tomas’ who recounted the same events using either positive or negative linguistic affect, thus, reflecting a positivity-offset or negativity-bias. After reading the assigned narrative, participants continued directly to the questions designed to assess the altruism felt toward Tomas. Participant altruism was assessed in two ways, via (1) 2AFC about whether or not they would offer Tomas help and (2) an open-ended question explaining their decision about whether or not to help Tomas. Response choice was evaluated in the former case and word choice was considered in the latter. Finally, participants completed the RLS (Reysen, 2005) and the PANAS (Watson et al., 1988).
The narratives followed other models of positive or negative language and priming (see Ito & Cacioppo, 2005; Mathews et al., 1995; Rozin & Royzman, 2010; Watson et al., 1988). The positive linguistic affect narrations included the positive words used in many of these studies. The words included in this list were interested, excited, strong, enthusiastic, proud, alert, attentive, happy, satisfied, good, cheerful, content, pleasant, sincere, and pure. The negative linguistic affect words used were guilty, bad, unattractive (“I do not like that”), stressed, disagreeable, depressed, sad, sorry, dirty, bad, scared, hostile, ashamed, and irritable (Ito & Cacioppo, 2005; Mathews et al., 1995; Rozin & Royzman, 2010; Watson et al., 1988;).
Four packets in each linguistic affect priming condition were not thoroughly completed and, thus, were not included in analyses. Given this, 132 completed test packets were analyzed for differences among tested variables between linguistic priming conditions. The significance level was set at .01 for all inferential analyses given the multiple dependent measures associated with a single priming instance.
Responses to the 2AFC altruistic decision question produced nominal count data, thus, the Chi-square test of association was used to test for differences in decisions (yes/no) between linguistic affect priming conditions. The Chi-square indicated no significant difference (p = .157) existed between linguistic affect priming conditions in this decision. Consequently, with respect to altruistic decision, we failed to reject the null hypothesis.
Participant responses to the open-ended question explaining their decision to help were typed into Microsoft Word, verbatim, from collected written responses and uploaded into the LIWC software (Tausczik & Pennebaker, 2009). The LIWC software assessed positive and negative word usage and provided a quantitative score for emotional valence of PA and NA. As LIWC output data was skewed to the left, word choice data was analyzed via the Mann-Whitney U (Mann & Whitney, 1957), the non-parametric alternative to the two-sample t-test. The Mann-Whitney U indicated participants in the positive linguistic affect priming condition used significantly more PA valence words in their explanation of why they choose to help Tomas than those in the negative condition (p = .005*). Additionally, participants in the negative linguistic affect priming condition used significantly more NA valence words than the positive linguistic affect condition (p = .002*). Analysis of the positive emotion words indicated a large effect size of 4.152. The effect size for the negative word output was even larger, sitting at 5.864 and both had a power of 1.0. Given this pattern of results, we reject the null hypothesis in favor of the alternative.
Likability data were tested for normality via the Anderson-Darling (Anderson & Darling, 1952) and both the sets associated with positive and negative linguistic affect were normally distributed. Therefore, a t-test was used to test for differences in Tomas’ rated likability between linguistic affect conditions. Analyses indicated positive Tomas was significantly more likable than the negative Tomas (p = < .001*). A large effect size was observed with Cohen-d = 0.857. Power analysis indicated strong results at 0.989. Given these results, we reject the null hypothesis in favor of the alternative (see Appendix B).
PA and NA scores were tested for normalcy using the Anderson-Darling (Anderson & Darling, 1952). PA data was normally distributed, but NA data was not. Therefore, PA was analyzed via two sample t-tests and no significant difference (p = .693) was observed by linguistic affect condition. Effect size for the PA data was small at 0.123 with a power of 0.031. NA data was analyzed with the Mann-Whitney U (Mann & Whitney, 1957) and no significant difference (p = .677) was found by linguistic affect condition. There was a moderate effect size of 0.507 with a power of 0.619 for this data. Therefore, we fail to reject the null hypothesis for this measure.
Limitations of Study 1
Participants read one long narrative and responded to all dependent measures following. The narrative might have been more strategically divided by survey measures to more directly examine dependent measures given the assigned priming condition. For this reason, a second study was employed to examine linguistic affect with only one dependent variable.
Study 2 – Using Emotion Words in a PA or NA Experience Narrative
Study 2 employed two levels of linguistic affect priming as the independent variable, however, this time participants wrote their own narratives and only affect was examined as the dependent variable. The research question sought to understand whether past experiences of PA and NA or the positive and negative emotion words used to describe those events were more predictive of overall affect.
- Hb5: The words used to describe events will more greatly impact affect than the type of event being described.
Convenience sampling was used to select 108 undergraduate participants from a small western liberal arts university. Participants who were provided informed consent were instructed they would be asked to write about either positive or negative experience using a randomly assigned list of words. Participants were informed they could withdraw from the study at any point without penalty and were provided research credit for their time. The purpose of this study was not explained to the participants to control positivity or negativity-biases. The researcher remained as neutral as possible during instructional phases. All data was collected and stored to promote and protect anonymity. To ensure objectivity, a double-blind research design was employed through the use of Microsoft Excel’s random number generator during assignment procedures.
Participants were given narrative packets which instructed participants to write about an experience of either a PA memory or an NA memory. The prompt for the PA condition was to write a narrative about, “a time when you felt active, alert, enthusiastic, happy, and engaged in pleasurable activities” and the prompt for the NA condition was to write a narrative about, “a time when you felt mood states of hostility, aversiveness, anger, contempt, guilt, fear, and nervousness” were present. These definitions of affect employed to design these prompts were adopted from Watson et al. (1988) and their studies examining PA and NA affect in the design of the PANAS.
Also included in these narrative packets was a list of either positive or negative emotion words participants were asked to use when writing their narrative. These words were the same as those employed in the first study. ‘Mad’ was used as an additional negative word so participants in each condition would have 15 words to choose from with the requirement they use at least 10 in their narrative. Participants circled the words they used to ensure they reached the minimum word use requirement.
Narrative packets, next, included two pages with blank lines to write their PA or NA experience narrative. Then, the PANAS was included to be finished upon completion of the written personal PA or NA narrative. PANAS words were counterbalanced to control for order effects from positive or negative words.
The PANAS gave participants both a PA and NA score. Total affect was examined by subtracting NA scores from PA scores sourced from the PANAS. This allowed us to examine the interaction of the words used and experiences written about in participant narratives with one dependent variable. A two-way ANOVA, general linear model, was run on the difference between PA and NA scores taking the provided word list into account. The significance level was set at 99%. The experiences written about proved not to be predictive of overall affect (p = .155), However, a significant main effect (p = .002*) did exist for the emotion words participants were instructed to use in their narrative. There was no interaction between experience and word condition (p = .190).
Power analysis of the non-significant data indicated there was a small effect size for both experience 0.019 and the interaction of experience and words 0.016 with observed power of 0.122 and 0.101 respectively. Regarding the statistically significant data for word usage, there was an effect size of 0.085 with an observed power of 0.684. This implies word usage is important in moderating affect. Ha5 was supported by this pattern of results (see Appendix C).
Further analysis of the data was conducted to understand this impact on PA and NA, separately. The purpose of conducting this analysis was supported by the significant results from the previous analysis and because Watson et al. (1988) indicated that PA and NA are distinct dimensions of emotion. Given this, analysis beyond subtracting the PA and NA scores to better understand affect was conducted.
A multivariate two-way MANOVA was conducted to understand the complex relationship between words, experience, PA, and NA. The significance level was set at 99%. Again, the experience written about was neither predictive of PA (p = .343) or NA (p = .189). Analysis indicated a small effect size for PA was 0.009 and NA 0.017. The observed power of experience data was 0.051 on PA and 0.102 for NA.
The MANOVA showed the emotion words used in the narratives did not predict PA (p = .035), but they did predict NA (p = .007*). Effect sizes for the words used in PA was 0.062 and was 0.069 for NA. The observed power for word usage was 0.319 and 0.559 for PA and NA respectively. This further analysis indicated that the words provided to the participants to use in their experiential narratives were more influential for regulating NA than PA.
Analysis of the interaction between words used and experiences written about produced non-significant results for PA (p = .076) and NA (p = .959). Effect size for this data was small for both PA 0.030 and NA < 0.001. The observed power for the interaction between the independent variables stood at 0.208 and 0.010 for PA and NA, respectively.
The results of these studies indicate the importance of positive or negative linguistic affect, as well as situations where linguistic affect did not operate. They can be extrapolated to undergraduate students and implicated in many contexts including the use of writing narratives to process positive and negative experiences. We also contend that, since the narrative in the first study was conversational in nature, our results regarding language use in a written piece may be extrapolated to verbal interactions with the presentation of linguistic affect within our environments.
Ha1, that positive linguistic affect priming would increase altruistic decisions, was not supported. Participants were willing to help a friend in need despite the language employed in the reading passage. Many participants in the negative linguistic affect condition stated they would Tomas solely because he was a friend, so results may have differed had Tomas been either a stranger or acquaintance.
Ha2, the hypothesis that those in the positive priming condition would use more PA valence words and those in the negative priming condition would use more PA valence words in their explanation of their helping decision, was supported. Positive or negative language can impact an individual’s word-choice when exposed to information in a reading passage or narrative. These results are interesting because of prior research indicating that word-priming may reverse individual positivity off-set or negativity-bias (Becker et al., 2015). Seligman (2006) describes the ability of individuals to cultivate learned optimism. The results of the current research may indicate that linguistic affect can run in a feedback loop with the presentation of positive or negative linguistic affect within our environment or social interactions impacting the way we describe that exposure. Interestingly, the same emotional valence of positive and negative words Tomas used to describe the events of his day were used to describe Tomas himself by participants in both conditions.
Ha3, the hypothesis that the individual using positive linguistic affect would be rated more likable than the one using negative, was also supported. The results of the current study indicate that positive linguistic affect makes an individual more likable than negative linguistic affect. This observation supports prior evidence that negative information is influential when making evaluative judgments of others (Martijn et al., 1992). The results of hypothesis testing for Ha2 and Ha3 support the view that an individual’s use of positive or negative linguistic affect not only impacts how they are perceived via likability, but also that the emotional valence of the language they use is how others tend to voice that perception of the individual.
Ha4, the hypothesis that positive or negative linguistic affect priming would impact participant affect, was not supported. This pattern of results may have been produced via study limitations. Specifically, an excess of time elapsed between linguistic affect priming and the dependent variable of the PANAS scale which was completed by participants at the end of the study.
The second study was employed to compensate for the shortcomings of the first. Only one dependent variable, affect, was examined in relation to the affective experience narrative written and the emotion words given to participants for use in the writing of their own narratives. The results of this second study indicate our word use when describing past experiences may moderate our affective states felt after the retelling of those events. This research supports the importance of linguistics in the field of psychology at large. It may be that our understanding of linguistic meanings help construct and predict our emotions (Barrett, 2017). This data also supports prior research that language priming may counteract a positivity-offset or negativity-bias (Becker et al., 2015; Crum, 2016).
Our choice of descriptive linguistic affect is contagious, at least when we are presented material through reading of a conversational narrative. This information is important in the modern world where words and information are transferred easily and quickly, sometimes even with just the press of a button on technological devices. Our linguistic affect also impacts our likability. Given this, it is recommended individuals take care with the type of information they share and the way in which they do so, as it not only influences others but also themselves.
Results of the second study showed the emotion words used to describe affective experiences are more influential to our affect than the type of experience being described. These results were especially notable in the NA experience condition. Participants who used positive words in both the negative experience and positive experience narratives had similar affect data. Those who used negative words to describe positive events had lower positive affect and higher negative affect than either of the two positive word groups, however, individuals assigned to use negative words to describe negative events had much higher negative affect and lower positive affect than those in all other conditions. This pattern of results supports our claim that linguistic affect moderates affect. In this case, we have a choice over the PA or NA we feel when we describe experiences to others. This research supports the need for heightened awareness and mindfulness of the language we use in daily interactions with others when describing events. These results are directly applicable in our lives. Linguistic affect not only impacts our own likability and the word-choice of those around us, but it also has the power to change our own affective state. This collection of studies highlights a need for greater conversational awareness and, perhaps, more positivity in social interactions.
Limitations of the first study, briefly mentioned earlier, centered on the number of dependent variables and lack of sufficient linguistic affect priming between measures. The number of variables was controlled for in the first study by setting the significance level at .01 rather than .05. Although there would be a need for a longer narrative and greater time commitment on the part of the participants, we recommend future research subdivide the independent variable of linguistic affect priming narrative into multiple parts that occur prior to each measure.
Another weakness of the first study was that participants could freely look back at the priming narrative while responding to all survey measures. This could explain the contagious nature of the emotional valence language in the priming narrative. Participants may have referenced the original narrative to remember what Tomas had said, however, this does not explain the participant’s choice in using similar PA or NA charged emotional valence words that Tomas used to explain their helping decisions. They could have simply said they did or did not like him for example, without using PA and NA emotion words. It was clear that the negative individual was much more unlikable in the analysis of likability data.
A weakness of the second study was that many, but not all, of the words used in the PANAS were the same as participants were instructed to use to describe affective experiences in their narratives. It is possible participants remembered the words they had used and, thus, rated them higher on the affect scale. Still, because previous research shows positive and negative word priming can impact cognitive biases (Becker et al., 2015; Crum II, 2016), we still find these results meaningful.
A limitation of both studies was that the sample consisted primarily of individuals from an affluent western institution located in a vacation town. It is possible individuals in this location may be more attached to positive language as the community culture tends to be more oriented toward positive experiences. Further research is necessary in more diverse populations to understand the role of language use within affect, specifically the meaning individuals attach to those words in different cultures.
The time of day may have also played a role of increased positivity, as research has indicated that the positivity-offset is more active early in the morning, stable during the day, and declines after evening (Watson et al., 1988). Data collection occurred in the morning and daytime, yet we observed higher than average NA means and lower PA means. Watson et al. (1998) found the PA mean among 660 American undergraduate students to be 29.7 with a NA mean of 14.8. The current study produced a mean PA of 27.56 and mean NA of 15.99 among undergraduate students. This may be an explanation to why we found insignificant data in altruistic decisions.
The current study only examined positive or negative emotion words with their direct meanings. It is recommended future studies in this area examine the effects of the same words, but with sarcastic meanings. Various cultures use language in a multitude of ways and sarcastic word use may not impact social desirability, PA, or NA. In fact, cultures with a high negative language sarcasm level who attach different meanings to words may show results opposite of those observed here, specifically regarding likability.
More research is needed on how linguistic affect impacts overall emotions. If variation in the construction of emotion exists (Barrett, 2017) then certainly it does across the different ways individuals use language and the meanings of the words they use to elicit affective states. The meanings we attach to words may be one of the direct lines in constructing and influencing our emotions in retellings of experiences. Future research might be completed in a variety of ways, considering human interactions rather than narrative reading and writing to study PA and NA states, as well as more complicated emotions across cultures is advised. Certain types of language could be triggering to individuals within specific populations. That same type of language could also be acceptable in other communities. The potential for research on linguistics within unique populations has a vast array of potential within the field. We also suggest future researchers examine physiological reactions such as heart rate or stress hormones in the context of linguistic affect.
Aknin, L. B., Dunn, E. W., & Norton, M. I. (2011). Happiness runs in a circular motion: Evidence for a positive feedback loop between prosocial spending and happiness. Journal of Happiness Studies, 13(2), 347-355.
Anderson, T. W., & Darling, D. A. (1952). Asymptomatic theory of certain goodness-of-fit criteria based on stochastic processes. Annals of Mathematical Statistics, 23, 193-212.
Baker, Z. G., Tou R. W. Y., Bryan, J. L., & Knee, R. (2017). Authenticity and well-being: Exploring positivity and negativity in interactions as a mediator. Personality and Individual Differences, 113, 235-239. https://doi.org/10.1016/j.paid.2017.03.018
Baolin, L., Zhixing, J., Zhonging, W., & Yu, H. (2010). The interaction between pictures and words: Evidence from positivity offset and negativity bias. Experimental Brain Research, 201(2), 141-153. http://dx.doi.org.unr.idm.oclc.org/10.1007/s00221-009-2018-8
Barrett, L. F. (2017). How emotions are made: The secret life of the brain. HMH Books.
Becker, E. S., Ferentzi, H., Ferrari, G., Möbius, M., Brugman, S., Custers, J., Geurtzen, N., Wouters, J., & Rinck, M. (2015). Always approach the bright side of life: A general positivity training reduces stress reactions in vulnerable individuals. Cognitive Therapy and Research, 40(1), 57-71. https://doi.org/10.1007/s10608-015-9716-2
Birmingham, W., Uchino, B. N., Smith, T. W., Light, K. C., & Sanbonmatsu, D. M. (2009). Social ties and cardiovascular function: An examination of relationship positivity and negativity during stress. International Journal of Psychophysiology, 74(2), 114-119. https://doi.org/10.1016/j.ijpsycho.2009.08.002
Bobić, J., Cvijetić, S., & Macan, J. (2015). Personality and self-perception of physical and emotional health among first-year university students. Drustvena Istrazivanja, 24(2), 219-237. http://dx.doi.org.unr.idm.oclc.org/10.5559/di.24.2.03
Brannon, L., Feist, J., & Updegraff, J. A. (2017). Health psychology: An introduction to behavior and health (9th ed.). Vancouver, B.C: Langara College.
Crum II, J. E. (2016). Effects of priming dialectic rational beliefs on irrational beliefs. Psi Chi Journal of Psychological Research, 21(1), 46-53. https://doi-org.unr.idm.oclc.org/10.24839/2164-8204.JN21.1.46
De Gucht, V., Fischler, B., & Heiser, W. (2004). Neuroticism, alexithymia, negative affect, and positive affect as determinants of medically unexplained symptoms. Personality and Individual Differences, 36(7), 1655-1667. https://doi.org/10.1016/j.paid.2003.06.012
Dizén, M., & Berenbaum, H. (2011). Cognitive correlates of emotional traits: Perceptions of self and others. Emotion, 11(1), 115-126. https://doi-org.unr.idm.oclc.org/10.1037/a0021415
Ekman, P. (2003).Emotions revealed: Recognizing faces and feelings to improve communication and emotional life.Times Books/Henry Holt and Co.
Freedman, J., & Combs, G. (1996). Narrative therapy: The social construct of preferred realities. W.W. Norton & Company.
Gibbons, H., Schnuerch, R., & Stahl, J. (2016). From positivity to negativity bias: Ambiguity affects the neurophysiological signatures of feedback processing. Journal of Cognitive Neuroscience, 28(4), 542-557. https://doi-org.unr.idm.oclc.org/10.1162/jocn_a_00921
Goodwin, R., & Engstrom, G. (2002). Personality and the perception of health in the general population. Psychological Medicine, 32(2), 325-332. http://dx.doi.org.unr.idm.oclc.org/10.1017/S0033291701005104
Herbert, C., Sfaerlea, A., & Blumenthal, T. (2013). Your emotion or mine: labeling feelings alters emotional face perception-an erp study on automatic and intentional affect labeling. Frontiers in Human Neuroscience. http://dx.doi.org.unr.idm.oclc.org/10.3389/fnhum.2013.00378
Ito, T., & Cacioppo, J. (2005). Variations on a human universal: Individual differences in positivity offset and negativity bias. Cognition & Emotion, 19(1), 1-26. https://doi-org.unr.idm.oclc.org/10.1080/02699930441000120
Lineweaver, T. T., & Brolsma, J. W. (2014). How you ask matters: An experimental investigation of the influence of mood on memory self-perceptions and their relationship with objective memory. Memory, 22(8), 1103-1115. https://doi-org.unr.idm.oclc.org/10.1080/09658211.2013.870209
Luengo Kanacri, B. P., Eisenberg, N., Thartori, E., Pastorelli, C., Uribe Tirado, L. M., Gerbino, M., & Caprara, G. V. (2017). Longitudinal relations among positivity, perceived positive school climate, and prosocial behavior in colombian adolescents. Child Development, 88(4), 1100-1114. https://doi-org.unr.idm.oclc.org/10.1111/cdev.12863
Machin, T. M., & Jeffries, C. H. (2017). Threat and opportunity: The impact of social inclusion and likability on anonymous feedback, self-esteem, and belonging. Personality and Individual Differences, 115, 1-6. https://doi-org.unr.idm.oclc.org/10.1016/j.paid.2016.11.055
Mann, H. B., & Whitney. D. R. (1947). On a test whether one of two random variables is stochastically larger than the other. Annals of Mathematical Statistics, 18, 50-60.
Martijn, C., Spears, R., van der Pligt, J., & Jakobs, E. (1992). Negativity and positivity effects in person perception and inference: Ability versus morality. European Journal of Social Psychology, 22(5), 453-463. https://doi-org.unr.idm.oclc.org/10.1002/ejsp.2420220504
Matthews, G., Pitcaithly, D., & Mann, R. L. E. (1995). Mood, neuroticism, and the encoding of affective words. Cognitive Therapy and Research, 19(5), 563-587.
Mayne, T. J., (1999). Negative affect and health: The importance of being earnest. Cognition & Emotion, 13(5), 601-635. https://doi-org.unr.idm.oclc.org/10.1080/026999399379203
Mendonça-de-Souza, A. C. F., Souza, G. G. L., Vieira, A., Fischer, N. L., Souza, W. F., Rumjanek, V. M., & Volchan, E. (2007). Negative affect as a predisposing factor for cortisol release after an acute stress—The impact of unpleasant priming. Stress: The International Journal on the Biology of Stress, 10(4), 362-367. https://doi-org.unr.idm.oclc.org/10.1080/10253890701379999
Reysen, S. (2005). Construction of a new scale: The Reysen Likability Scale. Social Behavior and Personality: An International Journal, 33(2), 201-208. https://doi.org/10.2224/sbp.2005.33.2.201
Rozin, P., Berman, L., & Royzman, E. (2010). Biases in use of positive and negative words across twenty natural languages. Cognition & Emotion, 24(3), 536-548. https://doi-org.unr.idm.oclc.org/10.1080/02699930902793462
Seligman, M. E. P. (2002). Authentic happiness: Using the new positive psychology to realize your potential for lasting fulfillment. Free Press.
Seligman, M. E. P. (2011). Flourish: A visionary new understanding of happiness and well-being. Atria Paperback.
Seligman, M. E. P. (2006). Learned optimism: How to change your mind and your life. Vintage Books.
Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology: An introduction. American Psychologist, 55(1), 5-14. http://dx.doi.org.unr.idm.oclc.org/10.1037/0003-066X.55.1.5
Small, J. (1989). Becoming naturally therapeutic: A return to the true essence of helping. Bantam.
Tausczik, Y. R., & Pennebaker, J. W. (2009). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24-54. https://doi-org.unr.idm.oclc.org/10.1177%2F0261927X09351676
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063-1070. http://dx.doi.org.unr.idm.oclc.org/10.1037/0022-3522.214.171.1243
Watson, D., & Pennebaker, J. W. (1989). Health complaints, stress, and distress: Exploring the central role of negative affectivity. Psychological Review, 96(2), 234-254. http://dx.doi.org.unr.idm.oclc.org/10.1037/0033-295X.96.2.234
Wenhui, L., Zhongqing, J., Ying, L., & Fei, D. (2016). ERP study of affirmative and negative sentences’ impact on self positivity bias. Journal of Psycholinguistic Research, 45(4), 769-778.