I have a copy of the paper Bostonian mentioned today.
The working paper version is at http://www.psy.cmu.edu/~siegler/Siegler-etal-inpressPsySci.pdf .
From the paper:
Marked individual and social class differences in mathematical knowledge are present
even in preschool and kindergarten (Case & Okamoto, 1996; Starkey, Klein, & Wakeley, 2004)
These differences are stable at least from kindergarten through fifth grade; children who start
ahead in mathematics generally stay ahead, and children who start behind generally stay behind.
((Duncan et al., 2007; Stevenson & Newman, 1986). Substantial correlations between early and
later knowledge are also present in other academic subjects, but differences among children in
mathematics knowledge are even more stable than in reading and other areas (Case, Griffin, &
Kelly, 1999; Duncan, et al., 2007).
*****************************************************
Fractions are worth teaching for their intrinsic importance, as ElizabethN wrote, but the paragraph above suggests a role for innate ability.