I think the key is the last clause, which suggests the author means a domain expert to be someone who has internalized important data patterns enough to quickly recognize them (or their absence) in a visualization. To get to that point, the expert must understand the normal relations between relavent variables so that unusual relations stand out.
Someone who's not an expert in the problem domain can spot patterns like correlation and outliers but not readily know if such patterns are important for the given variables.