Story ID
Part 2: Language Inquiry and Word Count Analysis
0.938.) Cross correlations among the NEO factors (both high and low across all five factors) are all also highly correlated (R2 ≥ 0.88).
Kucera-Francis Word Frequencies:
Finally, in examining the frequency distribution of cue words, it is important to look back to the selection criteria themselves. The final selection of cue words was based on a number of criteria (described in Chapter 2) including the balance of positive, neutral and negative concepts, concreteness, imageability, meaningfulness and arousal. In
addition, preference was given to cues with greater lexical frequency (as measured by the Kucera-Francis frequencies)222. The final distribution of cue words, however, includes words with frequencies ranging from 1 to 98, providing the opportunity to examine whether words that are more common in the English language elicit more moral
memories than those that appear less frequently. However, when these data are graphed Figure 35 demonstrates clearly that there is no relationship between the frequency of a word’s appearance in the lexicon and the number of memories that that cue elicited.
automated process. While automated text analysis of this sort cannot interpret complex moral narratives, it does provide an excellent way to quantify the overall structure of the narratives for comparison purposes. The LIWC analysis also allows us to compare our narratives to other samples in order to determine how typical they are. These data have been shown in Table 2 and demonstrate that the narratives are quite typical in their number of words per sentence, percentage of words found in the LIWC dictionary, number of words with greater than six letters and their overall grammatical content (as measured by function words).
LIWC2007 calculates the percentage of total text that is devoted to each of 80 variables. These categories include grammatical structures (e.g., nouns, verbs), personal motivators (e.g.. work, religion) and psychological constructs (e.g., affective, social and cognitive processes). The affective processes variables are particularly useful for our analyses because they allow us to compare the percentages of positive and negative emotion words in each narrative. The dictionary of words that connote positive emotions contains 406 words (e.g., love, sweet) and the dictionary of words that connote negative emotions contains 499 words (e.g., hurt, nasty). The dictionary of ‘negative words’ can be further broken down into anxiety, anger and sadness related words but since ‘positive words’ is not further subdivided, we have chosen not to use these divisions for our analyses.
LIWC Variable: Positive and Negative Emotion Words:
Using the LIWC text analysis, we examined the percentage of emotion words (both positive and negative) found in each narrative. The distribution of these data can be
seen in Figure 35. The average narrative in the database contained 2.7 ± 1.85% positive words and 2.0 ± 1.61% negative words, a significant difference (t[1514] = 7.10, p <
0.0001). This is a particularly surprising finding since there are more negative224 narratives in the database.
Additionally, Figure 35 shows that 7.9% (n = 60) of the narratives in the database contained no positive emotion words and 11.2% (n = 85) of the narratives contained no negative emotion words. 10 of these narratives overlap, hence they are entirely devoid of emotion words as found in the LIWC dictionaries225.
When analyzing the data, we noticed that the narratives with the highest
percentage of emotion words (both positive and negative) were particularly short. (The narrative with the highest percentage of positive emotion words (15.22%) was 92 words long and the narrative with the highest percentage of negative emotion words (11.59%) was 138 words long. The average word count for narratives was 217.6 words.)
Examination of these two narratives found that each participant repeated the cue word (or versions of the cue word) frequently226. We hypothesized that this might have happened whenever a participant had difficulty recalling a memory that fit the cue word, accounting for both the high emotional word count and the length of the narrative. In order to make sure that short narratives were not being disproportionately represented in the word count analyses, we inspected the data further. The word count of the 100 stories with the
highest percentage of positive emotional words is not significantly different (t[198] =
224 Negative narratives, in this case, are narratives that were cued by negative cue words.
225 It is important to point out that this is an important caveat about analyzing data based on LIWC tallies.
The program is only as accurate as its dictionaries are complete. The number of positive and negative emotion words is large (905 words combined) but it cannot contain all emotion words. More importantly, it cannot understand the nuance that may be imparted by the context of the narrative. However, it is this difficulty that is weighed against the difficulty of acquiring readers who can analyze each narrative individually.
226 These are memories 5588 and 5441.
1.82, p = 0.07) from the word count of the 100 stories with the lowest percentage of positive emotional words (including those stories with no positive emotional words).
Demographic Impact:
Similarly to the cue word analyses done in the previous section, we were interested in the influence that gender and neuropsychological measures might have on the number of emotional words in a narrative. While there is a wealth of literature227 to lead us to expect that men and women use emotion words with different frequencies, they are not statistically different for either positive emotional words (t[117] = 1.19, p = 0.24) or negative emotional words (t[168] = 0.15, p = 0.88). (These comparisons and those following were done by selecting the narratives with no emotional words (n = 60 for positive and n = 85 for negative) and comparing them with the same number of narratives with the highest percentage of emotional words in each group. The outliers discussed above were included in these analyses.) This finding is quite advantageous for using the narratives as new stimuli. Since there are no underlying gender cues in the emotional word usage, studies involving gender can use the narratives interchangeably. For an example of this, see Section 4, later in this chapter.
Research by McFarland and Buehler has found that when subjects involved in autobiographical recall are ‘ruminating’ on their own mood, they tend to recall memories with similar emotions. However, if instead the subjects are ‘reflecting’ on their own mood, they tend to recall memories with dissimilar emotions228. These data suggest that
227 As one example: Brody, L. R. Hall, J.A. (2000). Gender, Emotion and Expression. Gender Handbook of Emotions. M. H.-J. Lewis, J.M. New York, Guilford Press: 338-347.
228 McFarland, C. Buehler, R. (1998). "The Impact of Negative Affect on Autobiographical Memory: The Role of Self-Focused Attention to Moods." Journal of Personality and Social Psychology 75(6): 1424-1440.
subjects’ mood may affect the emotional content of the narratives. The Positive and Negative Affect Schedule measures the general mood of a participant using two orthogonal scales: Positive Affect (PA) and Negative Affect (NA)229. Together, these scales provide a general picture of a participant’s mood. Comparison of PA and NA scores for the subjects who told the narratives with the greatest percentage of emotional words with those for the subjects whose narratives contained no emotional words showed no significant difference for either positive emotional words (for PA, t[117] = 0.14, p = 0.89; for NA, t[117] = 0.42, p = 0.68) or for negative emotional words (for PA, t[168] = 0.36, p = 0.72; for NA, t[168] = 1.81, p = 0.07).