Narrative Content Analysis: Cognitive and Behavioral Reponses
Three undergraduate research assistants read the 127 narratives and classified each humor attempt based on four dimensions. The research assistants, who were blind to the purposes of the study at the time they coded the narratives, were trained and then asked to assess (a) the type of humor attempted, (b) the goal of the humor attempt, (c) the behavior of the agent following the attempt, (d) the reaction of the target following the attempt, and (e) the attribution the agent made for the success or failure of the attempt. These coding categories were theoretically derived as manipulation checks (type and goal of humor) or as foci of the study (behavior and cognitive attributions).
However, some of the coding dimensions for each category were developed inductively. The second author read through all of the narratives to identify emergent coding dimensions for each code category (Miles and Huberman
1994; Strauss and Corbin
1998). These emergent codes were added to codes derived from the literature. The first author then applied the coding scheme to a subset of narratives and together the authors revised the coding dimensions before training the research assistants. Although the literature on attribution theory identifies the importance of internal versus external attributions, this inductive process allowed us to uncover that participants distinguished between two different types of external attributions (those related to the target of the humor and those related to the humor method).
The research assistants, who were blind to the purposes of the study, coded all five categories for each story. First, the research assistants were asked to assess the type of humor attempted, i.e., whether the attempt made was positive or negative. Negative humor was defined as situations in which the agent attempted to be funny by putting down the targets, teasing them, or using sarcasm. For example, one participant wrote that he or she, “mocked (his or her friend), in a voice imitating her right after she said something stupid (ID# 1384).” This was coded as negative humor. Although these attempts were negative in that they included some level of provocation or criticism, they were intended to be humorous, to elicit positive affect, i.e., “to make the target laugh.” Even acts such as teasing, which can be used aggressively, also can be used to elicit amusement (Keltner et al.
2001). For example, Keltner et al. (
2001, p. 234) note that when teasing contains “provocations
accompanied by numerous off-record markers” such as hints, questions, rhetorical questions, or metaphors, it “will be perceived as playful.”
We coded for the type of humor (positive or negative) to see whether differences in the type of humor events individuals were likely to recall in the successful versus failed condition explained any of our findings about failed humor. We surmised that because teasing and other negative humor strategies can be perceived as aggressive when subtleties such as off-record markers are not used properly (Keltner et al.
2001), individuals may be more likely to fail when they use negative strategies than when they employ more positive types of humor. In our sample, 42 participants reported a negative humor attempts and 85 reported a non-negative humor attempts (
n = 127).
Second, the research assistants were asked to judge the goal of the humor attempt, i.e., whether the attempt entailed trying to increase the target’s positive affective bonds of friendship with the agent or trying to increase the target’s general positive affect. For example, someone with a goal of increasing positive affective bonds of friendship wrote that he or she attempted to be funny because he or she “wanted to maintain our friendship since we don’t get to see each other that much (ID# 1220).” Participants who tried to change the target’s positive affect made statements such as they were “attempting to use humor to make him feel better after a stressful day” (ID# 1227), “trying to put him in a better mood” (#1337) or trying “…to make him laugh and feel better (1220).” We coded for the goal of the humor attempt to confirm our underlying assumption that humor was being used to regulate the feelings of others. In our sample, 99 participants had the goals of “increasing positive affect,” 20 had the goal of “increasing affective bonds” and eight were coded as other.
Third, all narratives were coded for the behavioral response of the agent, i.e., whether, following the attempt, the agent laughed, made a new and different attempt at humor, repeated the same attempt, apologized for the attempt or did nothing. For example, an agent making a new different attempt stated that after the attempt they “followed up by making another joke to try to get [the target] to laugh again (ID #1240).” Conversely, someone who repeated the same attempt stated, “I continued telling the same story, but tried to make it more ridiculous so she would laugh (ID #1335).” One participant that was coded as apologizing stated, “Two of the people at the table chuckled but said that the joke was mean spirited. I apologized… (ID# 1307).” Finally, one participant who was coded as doing nothing after the attempt stated that after the attempt they “just didn’t make any more comments (ID #1337).” We coded for the agent’s behavior following the attempt to uncover patterns of behaviors associated with failed humor.
Fourth, the research assistants coded each narrative for the behavioral response of the target. Target’s behavioral responses were initially coded into two categories: responded as expected versus unexpected. These categories were coded by all three research assistants. To gain additional insight into the type of expected and unexpected responses that targets made, one research assistant and one coauthor went back through the narratives. The three categories that emerged from the data included the target laughed, the target was bothered by the attempt, and the target continued the interaction. Specifically, targets were only coded as laughing if this was explicitly mentioned in the response. For example, one person recalled, “(the target) laughed and thanked me for listening (ID# 1236).” Next, humor attempts were coded as bothersome when the target engaged in a specific negative behavior aimed at the agent. For example, one person recalled that the target, “certainly did not smile. They didn’t really look at me either. They just kept their eyes on the TV. Occasionally, they might say something curt, but it wasn’t anything that recognized my attempt to be fun and humorous (ID# 1211).” Finally, whether the target made an explicit attempt to continue the interaction was cataloged. For example, one person recalled that the target “smiled and told me an anecdote as well (ID# 1303).”
Finally, the research assistants coded participants’ cognitive attributions for the outcome of the humor attempt, i.e., whether the outcome of the attempt was due to themselves (an internal attribution), the target (an external attribution), or the method of humor attempt (an external attribution). Attributing the success or failure of the attempt to the agent was characterized by participants explicitly stating that something inherent in them had led to the success or failure of the attempt. For example, one participant said, “I was quite funny (ID# 1237),” whereas another said a joke failed because, “I have a different sense of humor from other people (ID# 1247).” Both of these examples were coded as internal self-attributions. Conversely, participants who attributed the success or failure of an attempt to target(s) made comments such as, “It failed because people are too stressed (ID# 1209),” and “my friends are dumb (ID# 1249).” Finally, those attributing failure to the attempt itself explicitly stated that something inherent in the attempt, as opposed to themselves, or the target, caused the outcome. For example, one participant stated, “The joke was inappropriate (ID# 1212).” We coded for the agent’s attributions following the attempt to uncover patterns of attributions associated with failed humor. All disagreements in classification were resolved by discussion among the research assistants. Finally, Shrout and Fleiss’ (1979) intraclass correlations (2, 1) were calculated to assess interrater reliability based on the premises that each judge rated their own random sample of targets and the judge’s data was combined for analysis. Interrater reliability exceeded 0.80 for all categories.