I think that more working memory is always better, in terms of overall intelligence as well as enabling higher achievement. While I certainly wouldn't disagree that one doesn't need a high working memory score on a particular test to be gifted, it's obvious to me that any particular gifted person would be more gifted with more working memory. I really don't think it's the case that a person gives exactly the same output, only slower, than they would with substantially more working memory.
People simply aren't computers, which can compute the exact same output regardless of underlying storage mechanisms, though the working memory concept apparently came in part from thinking about the brain as similar to a computer. And while I certainly think that a person can work around limitations by use of external aids like paper, and highly effective people would almost all tend to do this to some extent, it just can't be true that a person's brain operates extended onto paper the way it does natively.
However, the more I look into this the more I feel that working memory probably isn't tested well at this point. I also don't think working memory is just short-term memory for recently input symbols, etc. I looked at the technical reports for the WISC out of curiosity as to the justification for considering memory for digit strings etc. to be a good measure for working memory overall, and found a major reliance on a study of performance by schizophrenics on a card-based test, with little further explanation. I looked up that study (it's available as a PDF) and didn't come away with an understanding of why (for example) digit span testing would be a good measure of overall working memory.
With evidence that working memory tests give very different results depending on the subject matter, I'd want to know much more before concluding that a low working memory score on a particular IQ test was an indication of overall working memory performance. I think that first-hand accounts of people manipulating complex models in their heads indicate good working memory of a certain type, regardless of subtest scores. This means to me that the working memory scores on IQ tests are best seen as possible indicators of learning disabilities, but are useless for determining giftedness (though I still think working memory in general is important to giftedness).
One question of mine is the extent to which one's memory characteristics guide one's choice of activities (this idea is ripped off from another recent thread). Engineers might have working memory highly tuned for visual spatial manipulations, etc. Another question of mine is the extent to which certain functions or parts of working memory deal with retrieval and manipulation of information from long-term memory, and which other ones-- which might tend to be the ones more affected by learning disabilities on testing-- deal more with external input. Or, if that's a stupid question, what measures might be taken to minimize the impact of learning disabilities on input during working memory testing?
I think it would be interesting and very valuable to test working memory regarding a problem domain with which the testee was very familiar. That sort of testing might never show up on a general IQ test, as it's too individual-specific. However, I'd bet that even though working memory can highly influence intelligence (I'm convinced of this) any domain-specific working memory effects would tend to show up more on achievement tests, especially timed ones, except of course that they'd not be obvious.
