You have to customize your scale based on what you want to study. Self-efficacy is a broad topic and is influenced by a variety of factors.

Albert Bandura argued that generalized self-efficacy is too broad. In fact, if you dive into his work you'll see that he argues that in order to predict behavior from social-cognitive variables, such as self-efficacy, you have to have precision in what you're measuring - it should not be too broad or too narrow. Think of it this way:

If you wanted to determine whether self-efficacy predicted physical activity initiation, you would need to operationalize all the necessary constructs if you want to develop a reliable and valid means of prediction. So a more precise question might be, does task and/or barrier self-efficacy predict initiation and maintenance of leisure time walking? (task and barrier self-efficacy are two types of self-efficacy, sometimes also called "action" and "coping" self-efficacy, respectively). 

Now to measure these things, you might use pedometer for walking, you would have to define what constitutes initiation and maintenance, but what do you do to measure self-efficacy? If you use general self-efficacy, you're going to capture a lot of extra noise, because someone might have high general self-efficacy, but be unable to effectively initiate and maintain a walking regimen. If you use too narrow a measure, such as whether someone has high self-efficacy to acquire walking shoes, tie their shoe laces, walk, etc., this information may not really be useful either (well... as an aside, it might if you're talking about barriers to walking such as no access to safe walking areas, but I digress).

So, back on point, Bandura would argue that appropriate measurement would be walking self-efficacy, or exercise self-efficacy, or whatever you want to call it, but basically you would be measuring one's confidence to engage in the task (such as in task self-efficacy) and one's confidence to engage in the task in the face of barriers (such as in barrier self-efficacy). So the questions would have to tap the degree of self-efficacy as close to the measured behavior as possible, without being too narrow or too broad (such as in general self-efficacy).

Interestingly, Ralf Schwarzer has quite a bit written on general self-efficacy, but his model, the HAPA, doesn't really rely on this - it uses more specific types of self-efficacy and most of his published peer-reviewed papers seem to reflect the specificity that Bandura talks about, so I'm not sure why Dr. Schwarzer continues (?) to support the notion of general self-efficacy. I know there is quite a camp out there that still argues in support of GSE, but I dunno... I guess I'll accept the utility of General Self-Efficacy when someone can prove to me that EMDR isn't simply exposure therapy with wiggly (and unnecessary btw) eyes and hands...

If you want to read some of Banduras points Mr. Bonner is mentioning - i attached a file.

Thanks Timo - great find! The section on "

Domain Specification and Conceptual Analysis of

Self-Efficacy Multicausality" addresses what I was mentioning in my post. Bandura disucsses it in other writings as well.

I would not judge the GSE as hard as Mr. Bonner but i like to use a specific scale as well with the GSE, i.e. GSE + occupational self efficacy to see how close they are and if they are differences.