Excited I was to find this conference paper! "Yes," I thought! This will be fascinating. And then, after I finished reading all two pages, I felt disappointed. Is this the authors' fault? A bit. Is it my fault? A bit. You see, I sadly lack the necessary statistical literacy, or my statistical chops are seriously gummed up, to fully make sense of the results section. So, there's that.
Their conclusion, however, reads thus:
"An eye tracking study was presented in which several different
captioning styles were examined. Significant differences were
found between eye movement metrics depending on the
captioning style used, suggesting that captioning styles play an
important role in viewing strategies. Participants underwent large
amounts of saccadic crossovers and spent much less time reading
the captions when captions changed position frequently. Future
work is needed to fully examine the implications of these
differences" (emphasis added, p. 190).
This makes quite a bit of sense, especially when you consider that they tried four approaches to presenting the captions. (Read the article, heh!) Most notably they tried the traditional captioning positions as well as placing captions above the speakers when present on screen. If not on screen, the captions would be at the bottom. This left me wondering a couple things.
While it may be useful for comprehension to avoid lots of extra or overlapping eye movement, might it not be possible to have the caption placement near speakers occur during intense dialogue and conversation and then shift to traditional (at the bottom) placement when conversation off-screen alternates with significant NSIs (non-speech information)? That might be an interesting approach to captions to test out--especially in dialogue heavy video or film.
As to the article, I am grateful that the authors conducted and shared the research. I just wish the findings would have been more explicitly stated. Then again, there might not have been enough information, or data, to support broader generalizations or suggestions for practice. I respect that. However, in the interest of testing out other approaches to captioning, it would be nice to have some research-driven data from which to launch.