It is very difficult to get an objective scale for computer proficiency in one question. I have tried to do it for the purposes of screening out advanced computer users from usability tests, but not for your purposes of determining the relationship between proficiency and task completion. I don't think the response to a single answer will allow you to do this; I think the best approach for that is aomedia and ebeth's ideas for a set of task-related questions, from which you can try to define a proficiency score.
The way I would do this is come up with a list of tasks that would be the most relevant for your own test, and ask some variation of the following questions, give them 1-pt per question and add up the scores. Make sure the tasks either represent genuinely different levels of proficiency from each other, or represent different types of work, otherwise the correlation between tasks will catapault some users disproportionately higher relative to their actual proficiency.
When was the last time you did [task], if ever?
- Today or yesterday
- In the last week <--award 1 point for this answer or higher
- In the last month
How frequently do you [do task], if at all?
- Several times per week <- 1 pt
For what it's worth, I have found variations on the following theme of "how do others see you" effective for screening out most highly proficient users. It's highly subjective and varies from person to person, but it was good enough for my purposes.
Do your friends and family ask for your advice or help in using computers?
Do your co-workers see you as an expert user of [software]?