This shouldn't be a big surprise, but if it's someone you know or who's dear to you, it may still be surprising.
Ordinary deductive reasoning (if p, then q. p, therefore q) seems to a a preferential left-hemisphere ("left brain") type activity. But, if just a few words are changed and people are involved (e.g. If you give me sunflower seeds, the I'll give you poppy petals), then right hemisphere seems to get involved.
Of course we know of many smart people who may be so outstanding at reasoning in general, but then botch it up when it comes to reasoning with people. There are many reasons for this of course (sensitivity or memory to social cues, etc.), but it's also interesting to consider this aspect to reasoning within a social context.
Young children are notoriously bad at social reasoning, but this can be misleading at times if their non-social reasoning is strong (i.e. verbally gifted, but socially immature).
There are other common hazards of poor social reasoning - an especially common one is the tendency for young children to overestimate the intelligence or knowledge of others when they are unknowledgeable about something themselves. It's the idea that "Others know even if I don't know", which leads to all sorts of self-berating behavior ("I'm stupid, etc.).
It means we shouldn't assume that a child has a realistic notion of what others think or know. There are many normal developmental differences in social expertise, and explicit teaching can speed up the process if necessary.
Social Deductive Reasoning and fMRI