I'm not quite sure why we need to debate the issue of the relationship between numeracy and literacy.
I believe that across the spectrum of very literate to not very literate people, you will find a range of math performance. I'm not convinced by my experience or the research that there is a direct link, let along a correlation of any significance.
I suspect, as Martha provided, that we all have our "case studies", but none of mine are compelling enough to come to a conclusion. However, I also believe that we will continue to debate the issue because we need to, regardless of the "truth" of the matter.
We, as mathematicians, certainly don't want to take the position of no relationship between literacy and math because we would then be making ourselves suspect in the literary domain. Yikes...! And, for example, my daughter who has a Masters degree in creative writing is absolutely insufficient in standard math, yet she (not surprisingly) has very creative, intuitive and sometimes correct solutions for interesting algebraic kinds of problems.
I think that the argument/discussion/debate smacks of a lot of issues touched upon across almost all academic domains; would we posit that there is a relationship between cancer researchers and literacy, or do field anthropologists know math or literacy to a degree that would go beyond the criteria we propose for students who may be innumerate or illiterate?
It is fun to play with the idea and if it could be conclusively and repeatedly demonstrated and "proven" that a person who increases their mastery of math simultaneously increases their literacy performance, then I think getting them to master more math is a really nifty idea because their is collateral learning. However, would we dare to say that increasing one's literacy also increases math capability? Well, it might with word problems. I think that the recent brain-mapping research doesn't show that the math location lights up when reading poetry, and conversely... but then again, maybe we just don't yet know how to effectively measure either one ... later, mark