Assuming that recurrent IMI and musical obsessions lie on a continuum and that distressing IMI is associated with obsessive–compulsive symptoms, we aimed to cover a broad range of severity of IMI in our sample by including individuals with and without OCD. Individuals without a known diagnosis of OCDFootnote 1 (n = 291) were recruited as a convenience sample of students from the Department of Psychology and the Department of Music at Basel between September 2018 and June 2019. Exclusion criterion was age < 18 years.
Individuals with a lifetime diagnosis of OCDFootnote 2 (n = 81) were recruited between September 2019 and January 2020 for another study by our group (Wahl et al., 2021) in clinics specializing in cognitive behavior therapy treatment of OCD. The majority of participants (n = 57; 70.37%) completed the relevant questionnaires (see “Measures”) at the beginning of the other study, before any experimental testing had started. The remaining participants took part in the current study some time after they had completed the other study (range 54–489 days ago, M = 257.54, SD = 110.39). They had all agreed to be contacted for future research. All met DSM-IV (American Psychiatric Association, 2000) criteria for OCD, as assessed with the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I; First et al., 1997; Wittchen et al., 1997), at the time of inclusion in the other study. Severity of OCD at the time of inclusion in the other study was assessed with the Yale–Brown Obsessive–Compulsive Scale (Y-BOCS; Goodman et al., 1989; Hand & Büttner-Westphal, 1991). The SCID-I and Y-BOCS were deliberately not repeated for the current study to reduce the time burden for individuals with OCD. For the same reason, individuals with OCD completed only a subset of the questionnaires used in individuals without OCD. Participants received course credit (psychology students only), gift vouchers, or monetary reimbursement for participation.
Questionnaires to Assess Frequency, Interpretations, and Degree of Severity of IMI
The Characteristics of Earworms Questionnaire (CEAR) was devised to investigate several aspects of recurrent IMI for which no standardized questionnaire exists. It starts with a definition of recurrent IMI and continues to assess general descriptors (e.g., frequency of IMI), interpretations, and other responses to IMI. Some items are modified versions of the Cognitive Intrusions Questionnaire (Freeston et al., 1991) and the Revised Obsessive Intrusions Inventory (ROII; Purdon & Clark, 1994). In this study, we used the CEAR to assess frequency and interpretations of recurrent IMI (please see the Online Supplement for the CEAR).
Frequency of recurrent IMI (CEAR Frequency) was assessed with the question, “How often does intrusive music enter your mind?” Answers were provided on a 9-point scale (1 = once a month or less, 3 = at least once a week, 5 = at least once a day, 7 = at least 5 times a day, and 9 = at least 20 times a day). Convergent validity of this measure (and others) was assessed by correlating measure scores with each other. These correlations are presented in Table 1. CEAR Frequency showed a high correlation with frequency ratings on the IMIS (Floridou et al., 2015), which can be considered an indicator of high convergent validity.
Dysfunctional interpretations of recurrent IMI (CEAR Interpretations) were assessed with six items on the CEAR. Three of these items were devised according to descriptions of dysfunctional interpretations of recurrent IMI in the literature (Rafin, 2016; Taylor et al., 2014), for example, “During the past week, to what extent did the sudden occurrence of the intrusive music make you think that you may not be able to stop thinking about it?” Three additional items were modifications of ROII (Purdon & Clark, 1994) items, for example, “During the past week to what extent did experiencing repetitive, intrusive music signal to you that you were going crazy?” Participants answered each item on a Likert-type scale (for details see the Online Supplement). To obtain a total score, the mean of the six items was calculated.
Positive correlations of CEAR Interpretations with the three OBQ subscales of small to medium size in individuals without OCD (Table 1) can be considered initial indicators of convergent validity. The internal consistency was good in individuals with and without OCD and the total sample (Table 2).
The severity of recurrent IMI, operationalized as the extent to which IMI meets criteria for obsessions defined by the DSM-5, was assessed with the self-developed Severity of Intrusive Musical Imagery Scale (SIMS): four items assessing the characteristics of obsessions defined by Criterion A1 in the DSM-5 (intrusive unwanted thoughts that occur repeatedly and persistently, causing marked anxiety or distress), one item assessing responses to the IMI according to Criterion A2 (attempts to ignore, suppress, or neutralize IMI), and one item each assessing the three impairing consequences of IMI defined in Criterion B (engagement with IMI lasts longer than 1 h/day; causes clinically significant distress; causes social, occupational, or other functional impairments; for a complete version of the SIMS, see the Online Supplement). Items were answered with yes or no. To obtain a total score, the sum across all items was calculated as an index of severity of recurrent IMI (main outcome). Correlations with the four subscales of the IMIS were taken as indicators of convergent and discriminant validity. As expected, the SIMS correlated positively with the Negative Valence subscale of the IMIS. The size of the association was small to medium and can be considered a preliminary indicator of convergent validity. Correlations with the remaining subscales of the IMIS were small or nonexistent, which can be considered a preliminary indicator of discriminant validity (for all correlations, see Table 1). Additionally, the SIMS showed a small positive association with severity of obsessions (OCI-R), which is consistent with the intended overlap in the constructs of the SIMS and obsessions. Finally, in individuals without OCD, the SIMS was also positively associated with severity of depressive symptoms. This indicates that the SIMS constructs conceptually overlap with depressive symptoms in individuals without OCD, but not in individuals with OCD. The internal consistency of the SIMS was acceptable in the total sample and in individuals without OCD and high in individuals with OCD (Table 2).
As an additional variable, we calculated if participants reported meeting criteria for musical obsessions as defined by Taylor et al. (2014). The dichotomous variable “DSM-5 criteria for obsessions” took the value 1 if participants answered yes to all questions referring to Criterion A1 (intrusive unwanted thoughts that occur repeatedly and persistently, causing marked anxiety or distress), Criterion A2 (attempts to ignore, suppress, or neutralize IMI), and Criterion B (engagement with IMI lasts longer than 1 h/day or causes clinically significant distress or causes social, occupational, or other functional impairments) on the SIMS and the value 0 if they answered no to any question.
Obsessive–compulsive symptom severity was assessed with the Obsessive–Compulsive Inventory-Revised (OCI-R; Foa et al., 2002; Gönner et al., 2008) and depressive symptom severity with the Beck Depression Inventory-II (BDI-II; Beck et al., 1996; Hautzinger et al., 2006). Both are widely used measures with good validity and reliability. The OCI-R and BDI-II were used to provide preliminary indicators of convergent and discriminant validity of the SIMS.
The Involuntary Musical Imagery Scale (IMIS; Floridou et al., 2015) assesses involuntary musical imagery with four subscales. The Negative Valence subscale assesses negative consequences of IMI such as intensive efforts to stop IMI (e.g., “I try to block it”), emotional consequences such as worry or irritation, and general negative evaluations of IMI (e.g., “The experience of my earworms is unpleasant”). The Movement subscale assesses body movement that matches IMI rhythms. The Personal Reflections subscale assesses the interpretation of IMI as being related to unresolved personal issues or concerns, and the Help subscale assesses the perceived usefulness of IMI (Floridou et al., 2015). The IMIS has shown good reliability and validity (Cotter & Silvia, 2017; Floridou et al., 2015). In our study, the IMIS was used to provide preliminary indicators of convergent and discriminant validity of the SIMS and CEAR Frequency.
The Obsessive–Compulsive Beliefs Questionnaire (OBQ; Ertle et al., 2008) assesses three dysfunctional belief domains relevant to the development and maintenance of OCD: (a) importance and control of thoughts, (b) perfectionism and intolerance of uncertainty, and (c) inflated responsibility and overestimation of threat. It was included to provide an indication of the validity of CEAR Interpretations. The OBQ has high reliability and adequate validity (Ertle et al., 2008). It was explicitly conceptualized as a dimensional measure (Obsessive Compulsive Cognitions Working Group, 1997) and therefore is also applicable to individuals who have not been diagnosed with OCD. In the current study, the internal consistencies of all standardized measures were high to excellent with the exception of the IMIS Personal Reflections subscale, which showed acceptable consistency (Table 2). All measures were provided in German.
All participants provided written informed consent prior to participation. The study aim was described as investigating the associations between musical involvement, personal characteristics, and stress. Participants completed the questionnaires online via LimeSurvey (LimeSurvey Project, 2012). The first questionnaire collected demographic data (age, gender, years of education, and marital status) and was followed by questionnaires for individuals without OCD in the following order: the CEAR, the OBQ, the IMIS, the BDI-II, the SIMS, and the OCI-R. Individuals with OCD completed the CEAR, the BDI-II, the SIMS, and the OCI-R. The study was approved by the ethics committee of the Faculty of Psychology, University of Basel, 026–18-1, and by the ethics committee of North-West Switzerland, 2017–01,980.
Statistical Analysis and Models
Potential differences between the two samples in terms of sociodemographic data and the frequency (CEAR Frequency), interpretation (CEAR Interpretations), and severity (SIMS) of IMI were investigated using chi-square tests for the dichotomous variables (gender, years of education, marital status, musical obsessions [i.e., self-reported DSM-5 criteria]) and t tests for independent samples for the remaining variables (age, CEAR Frequency, CEAR Interpretations, SIMS, depressive symptoms, obsessive–compulsive symptoms).
Zero-Order Associations Between Predictors and Outcome
Zero-order associations between the predictors (CEAR Frequency and CEAR Interpretations) and outcome (SIMS) were analyzed using Pearson product-moment correlations. Effect sizes were interpreted according to Cohen (1988).
Testing the Interaction Effect of Frequency and Interpretations on Severity of IMI
A quasi-Poisson model was used to analyze the data. Poisson models are typically used when the outcome is a count variable (such as the SIMS in our study) and when the distribution is skewed right, especially if the expected value of its parameter (λ) is small. Note that the distribution of the SIMS was indeed skewed right (Fig. 1). Since count data often exhibit greater variance than expected from a Poisson distribution, the resulting overdispersion must be accounted for. Since our data were overdispersed, we used a quasi-Poisson model, in which standard errors are increased according to the amount of overdispersion relative to the ordinary Poisson model. The predictors were transformed since their frequency distributions were both skewed (log transformation for CEAR Interpretations and square root transformation for CEAR Frequency) to avoid undue influence of outliers on our model, using Cook’s distance as measure of influence (Fox, 2016).
We specified a model with severity of recurrent IMI (SIMS) as outcome, and frequency (CEAR Frequency) and dysfunctional interpretations (CEAR Interpretations) as predictors, including their interaction. Thus, we tested whether recurrent IMI was particularly high if both frequency and interpretation were elevated. Since the sample (with or without OCD) might conceivably have an influence on the results, the sample membership (OCD vs controls) was controlled for in the model. We have focused on the overall interaction in this initial testing of the two-factor model and depict the suggested moderations graphically, since the two-factor model of musical obsessions (Taylor et al., 2014) does not state explicitly which is the moderating variable. Significance level α was set to 5%.