According to current theories of reading, the reader's saccades are gu
ided primarily by spaces between words, clearly the most prominent vis
ual feature in most modern texts. This belief was investigated by reco
rding eye movements with unprecedented accuracy and precision while su
bjects read spaced and unspaced passages both silently and aloud. Mode
st increases in fixation durations and decreases in overall reading sp
eed were observed when unspaced texts were read. However, subjects rea
d unspaced texts with the same level of comprehension and percentage o
f regressions as they read spaced texts. The only global eye movement
parameter that changed appreciably when spaces were removed was progre
ssive (rightward) saccade length. Progressive saccades were shorter in
unspaced texts. However, unspaced texts were denser and narrower beca
use they were constructed so as to contain the same number of words/li
ne as the spaced texts. This meant that unspaced texts contained more
informational characters/degree of visual angle. The observed decrease
in progressive saccade length tended to be proportional to this incre
ase in text density. Therefore, the number of saccades/line of text re
mained approximately the same in both spaced and unspaced texts. Furth
ermore, a detailed examination of local eye movement properties, i.e.
where within words the subjects fixated and how many times they fixate
d words of different lengths, suggested that the same oculomotor strat
egy was used for reading spaced and unspaced texts. This was true for
both silent reading and reading aloud, Thus, a model that could explai
n reading spaced texts could also explain reading unspaced texts with
only a change of a single global parameter, namely, saccade length. We
conclude that the current tendency to emphasize spaces as guides to r
eading eye movements must be reconsidered. Words, not spaces, may serv
e as the perceptual units that guide the line of sight through the tex
t.