Studies using ER in the form of the detailed content analysis of dream reports have shown that most dream reports are non-affective (e.g., Domhoff, 2018; Hsu & Yu, 2016; Snyder, 1970), with less than one affective state per report (e.g., Hall & Van de Castle, 1966; Hsu & Yu, 2016). Also, such studies have demonstrated that dream reports contain more negative than positive affect (e.g., Brown & Donderi, 1986; Domhoff, 2018; Hall & Van de Castle, 1966; Hsu & Yu, 2016; Snyder, 1970), with anxiety and/or fear being the most prevalent discrete affective states (Hall & Van de Castle, 1966; Hsu & Yu, 2016; McCarley & Hoffman, 1981; Snyder, 1970).
In contrast, studies using SR have found the majority of dreams to be affective (Fosse et al., 2001; Foulkes et al., 1988; St-Onge, Lortie-Lussier, Mercier, Grenier, & De Koninck, 2005; Strauch & Meier, 1996; Yu, 2007), with several affective states per dream (Blick & Howe, 1984; Howe & Blick, 1983; Kahn & Hobson, 2002; Merritt et al., 1994; Nielsen et al., 1991). Results regarding affective valence are mixed: from the dominance of negative affect (Blick & Howe, 1984; Kahan & Claudatos, 2016; Merritt et al., 1994; Nielsen et al., 1991), to a balanced amount of positive and negative affect (Fosse et al., 2001; Kahn & Hobson, 2002; Yu, 2007) or affective tone (Blagrove et al., 2004; Strauch & Meier, 1996), to the dominance of positive affect (St-Onge et al., 2005). These inconsistencies arguably depend on additional methodological differences between the studies, such as the data collection environment and procedure (St-Onge et al., 2005), the number and type of discrete affect items included (Kahan & Claudatos, 2016; St-Onge et al., 2005), and individual differences such as age (Blick & Howe, 1984; St-Onge et al., 2005) and mental health status (Cartwright, Young, Mercer, & Bears, 1998). Which discrete affect is the most prevalent depends on which particular items have been measured. Joy and interest are typically the most frequent among the positive affective states, while anxiety, fear, and anger are the most prevalent among the negative affective states.
The findings reviewed above suggest that differences in the method of measurement—whether ER or SR has been used—may underlie (at least some) discrepancies in results. However, to determine whether this is indeed the case, systematic comparisons of SR and ER in the same sample are needed. In 1951, the eminent dream researcher Calvin S. Hall reported, albeit in a popular science journal, that dream reports content analysed by external judges contained more negative than positive affects, whereas the dreamers rated the same dreams to be more pleasant than unpleasant (Hall, 1951)16. More than two decades later, Stairs and Blick (1979) demonstrated that the participants’ own and judges’ ratings of dream affect had poor agreement. Another two decades later, Kahan and LaBerge (1996), who compared ER and SR of cognitive and metacognitive processes in dreams, reported a higher prevalence of dream affect with SR (93 %) than with ER (38 %).
Besides these early reports, prior to the studies conducted in this thesis, only one study had directly compared SR and ER of dream affect in the same participants. In 16 Specifically, Hall (1951) reported that, when written dream reports were content analysed,
64 % of all dream affects (what he termed emotions) were negative and only 18 % positive. But participants themselves rated the same dreams more often pleasant (41 %) than
this previous study, Schredl and Doll (1998) measured dream affect in 133 home dream reports using three different rating methods: (1) SR using two dimensional, unipolar (four-point Likert) scales, one measuring the intensity of positive affect and the other the intensity of negative affect in the dream as a whole; (2) ER based on the detailed content analysis of dream reports (ER-CA) using the Emotions scale from the Hall and Van de Castle (1966) content analysis system; and (3) ER based on the global ratings of the dream report as a whole (ER-GS) using the same two dimensional rating scales as for SR. The results showed that whereas with SR almost all of the dreams (99.2 %), with ER-CA less than half of the dream reports (42.1 %), were rated to contain affect. With ER-GS, however, more than two-thirds of the dream reports (86.5 %) were rated to contain affect. Thus, when judges could infer affective states from dream action and rate affect in the dream report as a whole, differences in the degree of dream affectivity between SR and ER were smaller, as opposed to when judges were instructed to rate only explicitly expressed affect. Nevertheless, ER (irrespective of whether content analysis or global rating scales were used) yielded more than twice as many negative than positive dream reports. With SR, the percentage of negatively and positively valenced dreams was more balanced, although negatively rated dreams still prevailed. Moreover, with SR, as compared with ER-GS, the intensities of both positive and negative affect were higher, but the differences were larger for positive affect. Although women expressed more affect in dream reports (i.e., with ER), their SR of affect did not differ from those of men. Also, there were no gender differences in the valence of dream affect.
Recently, Röver and Schredl (2017) replicated Schredl and Doll’s (1998) findings using SR and ER-GS. In addition, they found that longer dream reports and higher levels of extraversion were related to smaller differences between SR and ER- GS with respect to negative affect ratings. According to the authors, this suggests that a more detailed description of the dream enables external judges to more accurately rate dream affect. Extraverts are more likely to express affect, and therefore external raters may find it easier to detect affect in their reports. However, in the study, this only explained ratings of negative affect, whereas differences with regard to positive affect remained unexplained. Gender did not explain differences in the results obtained with the two measures.
In summary, ER and SR yield different results regarding the phenomenology of dream affect. This raises questions about the validity of the measures and calls for more systematic research on this methodological aspect. In particular, it is important to determine whether similar discrepancies are obtained when discrete affect is
measured with both ER and SR and when the number of positive and negative affect items is balanced, and whether these differences are specific to the data collection environment (i.e., home) or also emerge in other settings (i.e., in the sleep laboratory). To address these questions, Studies I and II were conducted.