Linguistic and perceptual-motor contributions to the kinematic properties of the braille reading finger

Abstract

Recordings of the dominant finger during the reading of braille sentences by experienced readers reveal that the velocity of the finger changes frequently during the traverse of a line of text. These changes, not previously reported, involve a multitude of accelerations and decelerations, as well as reversals of direction. We investigated the origin of these velocity intermittencies (as well as movement reversals) by asking readers to twice read out-loud or silently sentences comprising high- or low-frequency words which combined to make grammatical sentences that were either meaningful or nonmeaningful. In a control condition we asked braille readers to smoothly scan lines of braille comprised of meaningless cell combinations. Word frequency and re-reading each contribute to the kinematics of finger movements, but neither sentence meaning nor the mode of reading do so. The velocity intermittencies were so pervasive that they are not easily attributable either to linguistic processing, text familiarity, mode of reading, or to sensory–motor interactions with the textured patterns of braille, but seem integral to all braille finger movements except reversals. While language-related processing can affect the finger movements, the effects are superimposed on a highly intermittent velocity profile whose origin appears to lie in the motor control of slow movements.

Introduction

Relative to visual print reading, the linkages between the perceptual, cognitive, and motor processes of braille reading have not been deeply investigated. It is clear that visual reading and braille reading must share common types of operations: both can be held to involve perceptual information utilization, in letter, morpheme, and word recognition (with attendant demands on attention and memory), as well as the range of syntactic and semantic operations that give text a specific interpretation (Daneman, 1988, Foulke, 1991). In addition, both are highly dependent on concurrent active motor control: via fixations and saccadic eye movements or via finger movements. However, dissimilarities are also clear. As Bertelson, 1995, Millar, 1997 have pointed out, braille reading is held to begin with exhaustive and serial contact with the surface on which the text is embossed and to involve movements that are smoother than saccades and fixations.¹ As a consequence, empirically grounded models of braille reading need to show how the fingers move, how the interaction of skin and textured surface gives rise to pattern recognition, and if and how the linguistic processing of the braille code influences these perceptual and motor operations.

Since the finger movements of the braille reader need to be controlled with a degree of precision commensurate with the spatial details of the braille code, as well as the temporal rate at which words, syntax, and semantic content can be processed, analysis of the reading movements would benefit from measurements of higher order variables, such as velocity and acceleration. Consider Panel A of Fig. 1. When reading a line of text with a single dominant finger, the position of a braille reader’s dominant reading finger is highly linear when plotted as a function of time. That is, the dominant finger tends to move continuously from left to right, with contact maintained between skin and text surface. Similar position-by-time data have been reported previously (see, e.g., Bertelson et al., 1985, Millar, 1997, Mousty and Bertelson, 1985). Finger position-by-time plots suggest that braille reading is “smooth, with few variations in speed” (Bertelson, 1995, p. 94).

However, when the position data can be differentiated with respect to time, a measure of the instantaneous velocity of the finger as it crosses text becomes available (Fig. 1B). Such computations reveal that the finger rapidly alternates between phases of acceleration and deceleration and that there is considerable variability of velocity about the mean. Finger velocity tends to be neither constant nor smooth. Such an intermittent² profile is by no means peculiar to this trial; it is representative of all readings of all sentences by all individuals that we have recorded.

To what processes should such intermittencies be attributed? A priori, those exemplified in Fig. 1B can be attributed to more than one source. The fluctuations might originate by virtue of skin contact with textured surfaces – that is, in haptic texture perception. Raised dots of the braille code are, at this level, differentially rough surfaces that may well generate position-based fluctuations in the friction quotient. These fluctuations in resistance to movement across the surface might also impact the discharge of cutaneous receptors and proprioceptive afferents, and therefore motor efference, all of which characterize haptic explorations of texture (e.g., Hughes and Jansson, 1994, Katz, 1925–1989, Loomis and Lederman, 1986).

A second possibility is that, at the speeds characteristic of braille, the limb motor control system simply cannot generate either constant or smooth velocity profiles at the distal fingertips, even if that is the intent of readers. Intermittent velocity traces may be an emergent feature of finger movements that are slow enough to permit letter and word recognition but not fast enough to constitute singular trajectories.

It may be that the intermittency of the velocity profile is owing to fluctuations in the ongoing demands of linguistic comprehension. Do the kinematics of the movements of braille readers similarly reflect the instantaneous demands of text perception and processing in a manner that is comparable with print reading? Perhaps unfamiliar letter combinations, or rare words, or sentences of complex syntax, or unpredictable content have the effect of briefly reducing the velocity of the moving finger to accommodate longer lexical search, or parsing operations, or semantic processing. Psycholinguistic factors are known to have consistent effects on the patterns of eye fixations and saccades in visual reading (for recent reviews, see Kliegl et al., 2006, Rayner, 1998) and hence it would not surprise if there were analogous language processing modulations of reading movements in braille.

Our major objective was to investigate how each of these factors contributes to the kinematic profiles of the reading finger. First, we sought to establish the extent to which the kinematic characteristics of movements of the fingerpad over braille text are influenced by the texture composition of the braille cells, independently of needing to be read. We did this by asking readers to scan “as smoothly as possible” a single line of meaningless braille text comprised of repeated braille cells with zero, one, three, or five raised elements. We hypothesized that the velocity of scanning movements would be higher and the intermittency lower than during reading because there were no requirements to process text at any level. If the texture composition of braille cells influences the kinematics of finger movements, we expected to find text with more raised cells to create more resistance to smooth left–right movements and lines containing no raised cells at all to be scanned with the smoothest velocity profiles.

To measure text processing effects on finger kinematics, we created sentences that varied in two ways: by comprising either high- or low-frequency key words, which were embedded in sentences with or without meaning. If language processing demands are directly reflected in finger kinematics, then we should observe reduced mean velocities, more intermittent velocities, and more frequent reversals when reading sentences with low-frequency words. We examined these hypotheses in conjunction with two others. We asked whether second readings of sentences were faster and smoother than first readings, and whether the benefit of a second reading was greater for sentences with low-frequency words and/or sentences without semantic content. If first readings are slowed and made more intermittent by vagaries of lexical access or a lack of semantic context, is this true for second readings where memory for – and hence anticipation of – text is possible? In addition, we required readers to read either silently, so as to be able to repeat verbatim the contents of the sentences at completion of reading, or to read out-loud. These tasks differ primarily in their working memory demands (greater for silent reading) and their phonological encoding (greater for reading aloud but not absent in silent reading). Silent and oral reading have been found to differ in their visual reading characteristics: fixation durations tend to be longer and saccade lengths shorter in oral reading (see Rayner, 1984). Millar (1990) found that braille readers of all levels of experience were no faster on average reading aloud than silently. We sought a replication of this finding, while leaving open the possibility that the reading modes would have the different effects on finger movement kinematics, including velocity intermittency.