Your question is very broad. But from my reading of the literature, my hypothesis would be that there wouldn't be a difference in the degree to which learning curves are logistic.

In a very general sense, learning generally involves the accumulation of a vast number of smaller components. Some components are easier to acquire than others and some yield greater gain more rapidly. In many domains, processes like this presumably give rise to learning curves where the rate of learning declines monotonically with practice and approaches an asymptote. If the performance is operationalised as reaction time, then three parameter power ($y = a + b^(-cx)$) or exponential functions ($y = a + b*exp(-x)$) often provide good fit. See a discussion [here](http://acs.ist.psu.edu/papers/ritterS01.pdf).

If performance is operationalised as amount of knowledge, then the task is a little more difficult in terms of defining a natural metric of performance. 

In general, from the literature that I've read looking at differences between younger adults and older adults (e.g., see work by [Christopher Hertzog](http://www.psychology.gatech.edu/people/faculty/hertzog_christopher.php) or high intelligence and low intelligence individuals (e.g., see work by [Phillip Ackerman](http://www.psychology.gatech.edu/KanferAckerman/PhillipAckerman.html)), differences in learning curves tend to be expressed more in level and rate of learning. I.e., If there is a variable related to learning, then it tends to be correlated with initial and final performance. I.e., older adults and low intelligence individuals perform more poorly initially and after practice than younger adults and high intelligence individuals after equivalent amounts of practice.
In general, I don't see why the fundamental mechanisms of learning would be all that different at least as is expressed in the learning curves.

That said, most empirical papers that look at differences between groups, whether it be non-clinical versus clinical samples or any other group comparison often focus on overall performance. Furthermore, if you look at group level learning curves, the logistic function at the individual level could easily be smoothed out at the group level if the timing of the s-shape in the curve was different across individuals.