Once potential 'means' were identified, items were written for each subscale. This required making a number of decisions. Two prior measures developed based on the conceptualisation of control outlined by Skinner ( 1 995, 1996) were considered for use as a model. These two measures are the CAM! (Skinner, et. al., 1988) and SPOCQ (Wellborn,
1989). In accordance with these measures all items were written as statements. The major difference between the CAM! and the SPOCQ is the nature of the beliefs concerning the relation between the 'means' and the ends. The CAM! labels the relationship between the
'means' and the 'ends' as Means-Ends Beliefs and the items are worded in the third person. It asks about "general perceptions of the usefulness, utility, or importance of a given 'means' for achieving" a good outcome (Little & Lopez, 1997, p. 1 65). For
example, a 'powerful others' strategy item for the CAMI is "When kids do really good in school, is it because of the teacher?" (Skinner et al. , 1988). In contrast, the SPOCQ labels the relationship between the 'means' and the 'ends' as Strategy Beliefs and the items are worded in the first person. It asks about the perceptions concerning the usefulness, utility, and importance of a given 'means' for achieving a good outcome to the person. An example of a 'powerful others' strategy item for the SPOCQ is, "If I want to get good grades in a subject, I have to get along with my teacher" (Skinner 1995; Wellborn et al.,
1989).
Consideration was given to which of these approaches should be used in writing the items. It was decided to use the first person approach because it was thought that people' s beliefs about their own situation would have more impact on their feelings and behaviour. The third person approach may allow for some to endorse a 'means' as a source of control in general but not believe it is a source of control for themselves. It was also noted that
strategy beliefs are similar to locus of control beliefs. Both the measure by Levenson ( 1 972) and the measure by Wallston et al. ( 1 978) used the first person approach.
Because of the decision to use the fust person approach, the term 'Strategy Beliefs' was used in the present study to refer to items concerning the relationship between the 'means' and the 'ends' . To be consistent with this, the term 'Capacity Beliefs' was used in the present study to refer to items concerning the relation between the 'agent' and the
'means' . These terms are also consistent with those used by Skinner ( 1 995, 1 996) when she outlined the theory on which this measure is based.
In strict accordance with the definition of Capacity Beliefs (Skinner, 1 995, 1 996; Skinner et al., 1 990), items for the Capacity Belief scales were written to assess access to the
'means' rather than ability to bring about the 'means' . For example, to have a Family Capacity Belief it is not required that you believe that you can influence your family to help you. This is more like Bandura' s ( 1997) Proxy Control where the emphasis is still on personal influence. To have a Family Capacity Belief it is sufficient that you believe you have a family who is willing and able to help you. This maintains the theoretical move away from the focus on personal influence.
'Family' and 'Friends' were kept as separate 'means' because, although under the
construct of social support they are often conceptualised together, it was decided that they were different enough to warrant separate investigation. This came through strongly in the open-ended interviews where a number of people mentioned one of these 'means' but not the other. It was also felt that it would be easier to combine them rather than to separate them at a later stage if that was appropriate.
In writing the items several terms required extensive consideration. One of these was the term to use for 'powerful other' . In domain specific measures a specific term can be given to the powerful other. For example in the MHLC (Wallston et al., 1 978) the term 'doctor' is mainly used while the terms 'nurse' , 'family' , and friends' are also used. The same approach is used in the SPOCQ and CAMI (Skinner et.al., 1 988) where the term 'teacher' is used. However in the general domain the issue is not so clear. Levenson (1972) uses terms such as, 'powerful others' , 'those people above me' , 'powerful people' , 'those in positions of power' . These terms seemed very awkward, complex, and often like terms a
psychologist would use rather than real terms that people could relate to. Therefore it was decided that the term 'Boss' would be used. This is a term that all people are familiar with and clearly communicates what is intended. It was felt that while some people would not presently have a boss they could hopefully still be able to answer the questions based on past experience. While it is a term which is to some extent domain specific it was felt that in this circumstance it was better to be too specific than too general. The decision to use a specific term also made sense given that other areas of the 'powerful other' construct were covered through the 'Family' , 'Friends' and 'God' subscales
There was concern that the term 'God' might be seen as relating only to the Christian God or the Jewish God. Other terms were considered such as 'higher power' . Again it was felt that terms like 'higher power' were awkward and a number of people would not be able to relate to them. It might also be that the people from a Christian background (most of New Zealand) would not relate to this. In making this decision, consideration was given to the approach by the Religion in New Zealand study beliefs (Gendall et al., 1999). In this study, the term 'God' was used and seemed to cause no problems. This study also showed that over 60% of the New Zealand population indicated that they believed in God. From this it was concluded that a large proportion of the New Zealand population would understand the use of this term. After making this decision it was also discovered that the term 'God' was also used in a development of the God subscale of the Health Locus of Control Scale (Wallston et al., 1999).
As can be seen in Table 5 . 1 items were written for 22 sub scales. There were Capacity and Strategy Belief subscales for Ability, Effort, Luck, Boss, Family, Friends, God, Time, Money and Confidence. There was also a Control Beliefs sub scale and an Unknown Strategy subscale. Six items were written for each sub scale. These items were adaptations of the items in the SPOCQ. The adaptations removed the items from the education domain into the general domain.
Table 5 . 1
Subscales for MSOCS Capacity Beliefs Ability Capacity Effort Capacity Luck Capacity Boss Capacity Family Capacity Friends Capacity God Capacity Time Capacity Money Capacity Confidence Capacity Strategy Beliefs Ability Strategy Effort Strategy Luck Strategy Boss Strategy Family Strategy Friends Strategy God Strategy Time Strategy Money Strategy Confidence Strategy Unknown Strategy Control Beliefs
Once the items were written a small pilot study was conducted using a convenience sample of 33 people. (These items are contained in Appendix One). However, subsequently it was realised that the wording of the items within sub scales was too similar, which would restrict the coverage of the construct content area and could
artificially raise the internal reliability of the subscales (Cattell, Eber, & Tatsuoka, 1970).
To solve this problem more items were written to be as varied as possible while still
relating to the construct area. This process produced 8 to 1 2 items for each subscale
(except the Luck Capacity subscale which only had 4 items; Appendix Two).