Stiff, P. (1996). The end of the line: a survey of unjustified typography. Information Design Journal, 8(2), 125–152.
No empirical data, but a good overview. Science would tell us that inconsistent word-spacing as a result of justification may inhibit saccadic eye movement by creating irregular “jumps” for the eye to make.
I have not read a study that supports or refutes this.
Anecdotal wisdom from the field of typography would have us believe that large gaps between consecutive lines will create vertical “rivers” of white space which draw the eye downwards as opposed to leading it to the right when the gaps between the words become larger than the space between the lines.
I have not read a study that supports or refutes this.
Anecdotal wisdom from the field of typography would also have us believe that the irregular shape of the right edge of a block of text helps us orient our eye on the page, assisting us in our return saccade to the next line.
I have not read a study that supports or refutes this.
I did however conducted a small experiment using eye tracking equipment which showed that when reading justified text, return saccades were less less accurate when compared to those while reading text set flush-left text. However, when reading the justified text, the duration of the landing fixations, and the distance of the correction saccades required in order to continue reading were very similar from line to line when compared to the landing fixations and correction saccades as seen when reading text set flush-left.
I hypothesize that this is because — despite the presence of an irregular rag in a flush-left setting — the distance and trajectory of the return saccade when reading justified text remains constant from line to line resulting in a muscle-memory of sorts.
I need to collect significantly more data to support or refute this hypothesis (feel free to scoop me. It would make a great thesis).
Long/short, what is published in typography books is largely anecdotal wisdom, not supported by scientific research. And the bulk of the research that is out there is very new, requiring further exploration before claims of any confidence can be made. For example, there is still no agreement regarding the legibility of serif compared sans-serif typefaces.