Cue validity for a cue (i.e. feature) and a category is defined as the conditional probability that an item belongs to the category if it has the cue.
Cue validity for a category is defined as the sum of cue validities for all cues associated with the category.
According to the definition, the cue validity for a category is higher if by knowing an object's feature we can better predict the category.
We might interpret Rosch as saying that "superordinate categories have lower total cue validity," because "they have fewer common attributes." But this is not necessarily true.
For example, consider "chair" (basic-level category) and "furniture" (supordinate category). Certainly, "chair" has more family resemblance than "furniture", but does "chair" have higher cue validity?
It would seem that the cue validity for "furniture" is higher that it is for "chair" since knowing features such as seat, legs and wood increases the success of correctly saying that the item is "furniture" more than it increases the success of correctly saying that the item is a "chair." Consider that these features could also indicate a "stool" or "bench", but in all cases it is still correct to call it "furniture."
Of course, this is just one example and assumes specific features. It's not clear if this example applies to superordinate categories in general.