Measures

The Language Background Questionnaire (Thomas & Gathercole, 2007) is a self‐report measure of language proficiency and use. Questions determine age of language acquisition, perceived language ability (reading, writing, speaking, and understanding), and use of language in everyday life. We used this questionnaire to determine (1) age of language acquisition, (2) self‐reported proficiency, and (3) frequency of language usage. As a consequence of our convenience sampling method of recruiting voice‐hearers into the study, there were an uneven number of participants falling into age of acquisition groupings. The majority of participants were simultaneous bilinguals, having learned both Welsh and English from birth (N = 22). L1 Welsh bilinguals (N = 10) were classed as those who had learned Welsh from birth and begun to learn English later (between the ages of 4 and 8 years). The L1 English group (N = 5) included people who had learned English from birth and begun to learn Welsh later (between the ages of 8 and 14 years).

We calculated an overall subjective language proficiency score for both Welsh and English separately. Composite scores were computed as the average proficiency scores (1 = low proficiency to 5 = high proficiency) across dimensions of speaking, understanding, reading, and writing for both Welsh and English separately. Across the sample, there was greater proficiency in English (M = 4.86; SE = .06) compared to Welsh (M = 4.65, SE = .08), t(36) = −2.46, p = .019.

Based on the answers from the LBQ, we calculated a frequency of language use variable to determine the proportion of each language participants used throughout their lives. One section of the LBQ estimates the proportion of each language a person has used from childhood to adulthood with parents, siblings, teachers, and friends, for example, ‘What language would you speak with your mother in the home at the following ages (before school, primary school, secondary school, and as an adult)?’. The response options were Welsh, Mostly Welsh, Bilingual, Mostly English, and English. We collapsed the Welsh and Mostly Welsh into one category (Welsh) and English and Mostly English into one category (English). In each domain, a score of + 1 was awarded for reported Welsh language use; a score of 0 was given to equal language use (bilingual); and a score of −1 was given for English language use. These scores were then averaged across domains. Therefore, a more positive score indicates greater frequency of Welsh language, and a more negative score indicates greater frequency of English language usage.

The Mental Health Research Institute Unusual Perceptions Schedule (MUPS; Carter, Mackinnon, Howard, Zeegers, & Copolov, 1995) is a semi‐structured, phenomenological interview designed to be more flexible than standard, structured clinical interviews by providing sufficient freedom for interviewees to expand upon their voice‐hearing experiences. The MUPS is divided into several subsections ranging from the physical to the personal characteristics of voices, as well as cognitive processes, and perceptions of voices. The schedule consists of 365 questions. Reliability has been reported to be high between raters on separate sections of the MUPS (Ms = 0.81–0.98; see Carter et al., 1995). From the MUPS interviews, we extracted the language the voice/voices used to speak to the voice‐hearers.

The interview data were initially scored as a categorical variable for language of voices. For example, for the question, ‘In what language did your voice speak to you?’, voice‐hearers would typically report the language in which they experienced their voices. We subsequently asked the voice‐hearers to provide an estimate of how often they experienced their voices in each language, and categorized them into Welsh, Mostly Welsh, Bilingual, Mostly English, or English, based on a similar criteria to the LBQ.

We transformed these categories into a continuous variable called the voice language index to conduct further regression analyses alongside other continuous variables. For example, if a person heard a single voice only in English, they would get a score of −1; if they heard a bilingual voice, they would get a score of 0; and if they experienced a single Welsh voice, they would receive a score of +1. If a person had more than one voice, the language of all voices was assessed and the resulting language index score would be the amount of Welsh, English, or both that the voices spoke divided by the number of voices heard. For example, one participant experienced three voices. One of these voices spoke Welsh 100% of the time (+1), another spoke English 100% of the time (−1), and the final voice was bilingual (0) and alternated between English and Welsh depending on context; the resulting voice language index score was 0.

The Psychosis Screening Questionnaire (PSQ; Bebbington & Nayani, 1995) was used to verify the presence of psychosis‐like experiences on entry to the study. The PSQ is validated in terms of its specificity and sensitivity to the presence of psychosis‐like experiences, which are separated into five types consisting of Thought Interference, Persecution, Perceptual Abnormalities, Strange Experiences, and Hallucinations. Each section consists of an initial screening question, for example, ‘Over the past year, have you felt your thoughts were being directly interfered with or controlled by another person?’. If the initial screening question was answered affirmatively, participants would answer the follow‐on questions.