Translation
The translation procedures were informed by the European Research Group on Health Outcome recommendations [19] and the International Test Commission Guidelines [20]. The repeated forward-backward translation procedure was adopted as the most suitable strategy that was pragmatically possible.
In Phase 1, four married couples of British and Japanese origin were separately asked to translate the original scale into Japanese with each couple among themselves discussing the conceptual, semantic and content equivalence between the original and their translation. The four couples were selected in accordance with the following criteria:
(1) one member of the couple was a native English speaker and the other a native Japanese speaker;
(2) both members were reared and educated either in English in an English-speaking country or in Japanese in Japan until at least 18 years of age;
(3) they have spent more than five years together since they married.
These criteria were used to identify translators who were familiar with both their own language and cultural background and that of the alternative language. The use of married couples was based on the opportunity such couples presented for exchanging a native speaker's insight into expressions in different languages among an intimate couple without the bias of representativeness introduced by restricting translators to those with a formal academic training. None of the individuals involved were professional translators. Thus it was hoped that an equivalent translation would be produced that was potentially more representative of the wider cultures than would be gained from a bilingual person or highly trained translators. All four couples happened to be of a British male and a Japanese female. They were fully informed of the objectives of their role in the whole procedure and were asked to discuss conceptual, semantic and content equivalence and to emphasise meaning rather than word-to-word translation. One of the authors (CM whose first language is Japanese) unified the four Japanese translations created by this process into a single translated version. Selection among alternative Japanese translations was based upon the perceived "naturalness" of linguistic expression in the Japanese language version.
In Phase 2, an additional couple was identified using the same criteria. They were asked to back-translate the Japanese version produced in Phase 1 without sight of the original version. In Phase 3, five university lecturers at the authors' college (native English speakers) compared the original scale and the back-translation brought about by Phase 2, and checked for semantic discrepancies. In Phase 4, the author altered the Japanese expression of the parts found to be problematic in Phase 3 with reference to any alternatives rejected in Phase 1. An example of differing translation when put back into English was Item 7 "how often did you feel that things were going as you expected in this past month?" as opposed to the original statement "in the last month, how often have you felt that things were going your way?" All of the lecturers pointed out that the translation could be negative but the original was more positive, thus the Japanese translation was altered so that it did not include negative meanings.
The couple used in Phase 2 re-translated them into English. One of the panel used in Phase 3 checked discrepancies between the original scale and the re-translation. Detailed discussion of cultural difference and nuance aimed to ensure semantic equivalence and to overcome conceptual differences by identifying parallel concepts. This process was repeated until problems were resolved.
Equivalence assessment
Respondents
Data were collected in the UK using the original English language PSS and in Japan using the translated version which we refer to as the PSS-J. Participants were recruited from full-time BSc nursing and pharmacy students of all years (1 to 4) at single university institutions in central London and Tokyo. Non-native English/Japanese speakers were excluded as appropriate to the version of the scale being tested. Data were obtained from 131 nursing and 91 pharmacy students in the UK (n = 222) of whom 28 were male (12.6%) and 194 were female (87.4%). Ages ranged from 18 to 45 and the mean age was 22.1 (SD = 4.5). The Japanese sample comprised 344 nursing and 976 pharmacy students (n = 1320) of whom 296 were male (22.4%), 1018 were female (77.1%) and 6 (0.5%) did not indicate their gender. Ages ranged from 18 to 44 and the mean was 20.6 (SD = 2.8). The differences in sample size were largely dictated by the size of the student cohorts in each institution. The response rate was 70.3% in the UK and 83.6% in Japan.
Data Collection
The questionnaire was administered to the students in a class setting. After permission for access to the students was obtained from the head of department and the course leader, the investigator visited the class in a room before or after a lecture. The questionnaires were distributed only to students who agreed to participate in the study. For the pharmacy students in the UK, it was not possible for all students to complete the questionnaires immediately owing to their tight academic time schedule. Therefore, a designated box was allocated in their school, and they could choose to complete the questionnaire immediately or to return it in the box later. For all other groups questionnaires were gathered in the envelopes provided immediately after they finished completing the questionnaire in the room.
Prior to the data collection, a pilot study was conducted to check the feasibility of the administration process and the credibility of the original and translated instruments. The questionnaire was administered on nursing students undertaking their postgraduate courses in the same university as the main study (n = 38 in the UK, n = 23 in Japan). The procedure of data collection in the pilot studies was exactly the same as in the main study. No problem arose regarding the administration process, including in the data collection and data handling procedure in both countries. Respondents in the UK appeared sometimes to miss negative words such as "no," "not" and "unable" when reading the items and scored them in reverse. Thus, amendments were made to the questionnaire, emboldening and underlining negative words. As for the Japanese version of the questionnaire, no problem was observed, and therefore no amendment was made to it.
Data Analysis
Factor structure was assessed by using exploratory factor analysis. With principal component analysis, the largest two factors were extracted and subjected to Varimax rotation with Kaiser normalisation. For the purpose of establishing equivalence, a two-factor solution was used as this had been identified in the original PSS [21–24]. After the analysis was carried out, target rotation [25] was performed to estimate factorial agreement of the two factors of the PSS and PSS-J for the different culture groups, which determine the construct equivalence. Cronbach's alpha coefficient was calculated to examine internal consistency reliability for the PSS and PSS-J and for each factor of the two scales.