Let me pull a quote from the discussion section that sums things up nicely:
Visual political knowledge is different from verbal political knowledge and represents a previously un-measured element of political involvement. This study has shown that adding visuals to otherwise identical all-verbal knowledge questions significantly increases correct responses. This finding strongly suggests that some people with substantive knowledge of political figures respond incorrectly to knowledge questions about them just because they lack a phonological representation of the person (the politician’s name). Allowed to draw on a visual representation (the politician’s face), they are able to report accurate conceptual knowledge about the politician.For the entire population, the effect is relatively small. For some groups, though, it's larger. Women, for example, are superior to men at facial recognition. Perhaps verbal tests of political knowledge, which often place women lower than men in scores, are masking true knowledge differences because they do not include visual knowledge.