On the Nonsmooth, Nonconstant Velocity of Braille Reading and Reversals

Abstract

Relative to print reading, braille-reading finger movements are held to be of more constant speed, with continuous and exhaustive contact with all words. However, the continuity of movements is intermittent in two distinct ways: (a) readers reverse direction and reread material already encountered and (b) the continual fluctuations of velocity between phases of acceleration and deceleration. We analyze recordings of experienced readers’ encounters with ambiguous sentences predicted to encourage more reversals, and we undertake comparative kinematic analyses of the finger movements during such reading. The data suggest that when reversals are initiated and where they are targeted reflect language-processing demands. However, their kinematic properties are not qualitatively different from those of forward reading. We conclude that how readers of braille move the reading finger—both forward and in reverse—is primarily influenced by the control properties of low-velocity movements and only secondarily by language properties.

ACKNOWLEDGMENTS

The research was funded by the University of Auckland, Faculty of Science Research Development Fund. We thank the reader-participants for their enthusiasm, insights, and commitment to research on braille. We thank Jonathan McKeown-Green for many discussions, and Ronan Reilly and three anonymous reviewers for comments on an earlier version of the manuscript. Portions of the data were reported at the 19th Annual Conference of the Society for the Scientific Study of Reading, July 11 to 14, 2012, Montréal, Canada.

Notes

1. ¹Movement smoothness does not imply execution at a constant velocity. The smoothest movements are those with the fewest changes in phase from acceleration to deceleration. These movements are not executed at a constant velocity, however. In fact, the range of velocities a finger may go through can be extreme: accelerating to a peak, then decelerating to stop in the smoothest instance. Nor is it the case that movements ramp up to a set velocity, which can be sustained without fluctuation for a period of time and ramped down to zero.

2. ²In the print literature, reverse direction eye movements are referred to as saccadic regressions. Because we have little evidence that braille reversals have the same causes, goals, or execution characteristics as saccadic regressions, and because braille readers cover text at significantly slower rates that do print readers, we preliminarily refer to them as reversals.

3. ³Reversals are evident in both one- and two-handed reading. In the former, because all information is being extracted from stimulation from one or more adjacent fingerpads, it is easier to interpret the location from which the reversal is made. In the latter, the trailing finger(s) of the left hand will usually make the reversal, whereas those of the right will either mark the origin of the reversal or continue moving; see, for example, Bertelson et al. (1985) and Millar (1997). An observer’s difficulty is in knowing where attention is being directed in the latter case: The input from which finger(s) of which hand is being read?

4. ⁴We asked participants to read with one finger only, regardless of their preferred mode. This enabled us to measure with some confidence what sections of the material were being read and when. This would not have been the possible if we had allowed readers to use more than one finger (see Footnote 3).

5. ⁵This is not to say the readers are not intending to move smoothly or at a constant velocity. They may well be. We are arguing that, even if that was the intent, they do not do so.

6. ⁶It is entirely possible that mean velocity for garden-path sentences (relative to control sentences) was reduced because its computation included more reversals and hence more negative velocities. It is common practice in print reading research to measure separately first fixations and subsequent refixations that occur as a result of regressions. This is more problematic in braille reading for several reasons. It is not straightforward to measure and then subtract out the rereading time on a cell or word: There is no fixation duration. Furthermore, an area with a lateral dimension between 10 and 16 mm of skin studded with mechanoreceptors may capture the contents of more than one braille cell, which makes cell reading times difficult to measure accurately. In addition, our traces show that rereading is more rapid (see Figures 1 and 2) and this higher velocity carries over to subsequent words. Other methods of measuring braille reading, which sum durations over cells or words (e.g., Aranyanak & Reilly, 2012; Breidegard, Eriksson, Fellenius, Jönsson, Holmqvist & Strömqvist, 2008), may be useful in this regard.

7. ⁷Plots such as Figures 1 and 2, which show velocity as a function finger position on the line rather than as a function of time, are not appropriate to measure this. Separate analyses of velocity by time plots confirm the absence of any stoppages.

8. ⁸Inhoff et al. ( and 2), with a p2005) suggested that short-range regressions (such as to the immediately preceding word) are specified via precise representations of target locations but that longer range saccades (which are rare) may not be.