#AUTHOR: Bertelson, Paul; Mousty, Philippe and D'Alimonte, Graziella

#TITLE: A Study of Braille Reading: 2. Patterns of Hand Activity in One-Handed and Two-Handed Reading

#ORGANIZATION: Université libre de Bruxelles, Brussels, Belgium

#CATEGORY: Braille Reading, Hand Movement

#ABSTRACT: Investigates the patterns of activity of each hand in persons reading Braille materials.

A Study of Braille Reading:

2. Patterns of Hand Activity

in One-Handed and Two-Handed Reading

Paul Bertelson, Philippe Mousty and Graziella D'Alimonte

Université libre de Bruxelles, Brussels, Belgium

Video recordings of the hands of 24 blind adults reading aloud prose, statistical approximations and scrambled words with either hand alone and with two hands were analysed. Hand activity involved forward scanning movements, regressions and returns to the next line. In two- handed reading, most readers resorted to mixed movement patterns, where a first segment of the line was explored by the left hand alone, a second segment by the two hands conjointly and a third segment by the right hand alone, each hand returning to the line while the other one was reading. In subjects who dissociated the hands to a great extent, the left hand generally started scanning the new line before the right hand had reached the end of the preceding line. This simultaneous disjoint exploration occurred with little slowing down of forward scanning, and involved parallel collection of textual information by the two hands. There were large and reliable individual differences in the relative sizes of the segments explored using the different hand combinations. Hand dissociation was correlated positively with individual reading speed in both two-handed and one-handed reading, and reductions in the degree of contextual constraint induced the adoption of more conjoint patterns of exploration. Total reading time was analyzed into components corresponding to regressions, line transitions and forward scanning. The main part of the gain from two-handed reading resulted in most subjects from savings in line transition time, but the most conjoint readers gained also on forward scanning and on regressions. The lower overall speed of conjoint readers was due not only to longer transition times but also to slower forward scanning and more time spent regressing.

Introduction

Since the principal differences between tactile and visual reading lie presumably in the characteristics of the perceptual systems involved in text exploration, knowing how the hands of the reader move is an obvious prerequisite to an understanding of braille reading. Examination of the literature on braille (Foulke, 1982) shows, however, that patterns of finger exploration have received relatively little attention. Useful elements of information, based either on questioning readers (Bürklen, 1917) or teachers (Lowenfeld, Abel and Hatlen, 1969), on direct observation (Grasemann, 1917) or on various types of recordings (Holland and Fertsch Eatman, 1933; Fertsch, 1947; Kusajima, 1974; Davidson, Wiles-Kettenmann, Haber and Appelle, 1980) are available, but one finds no really systematic description.

One point that emerges relatively clearly is that reading is generally accomplished with the tips of the index fingers. The majority of the readers in the studies by Bürklen (1917) and by Lowenfeld et al.(1969) are described as using exclusively these fingers. The use of finger combinations involving the index and the middle or even ring fingers is reported in the latter for a minority of readers. However, as the authors - note, the data do not guarantee that these other fingers were used for "actual reading", since the observers were provided with no guidelines for deciding when a finger was reading and probably adopted various more or less implicit criteria. The results of Foulke (1964), who had subjects read prose with all fingers except the thumbs, show that reading is possible with each of them, but that the indexes are much faster.

Concerning patterns of hand movements, descriptions in terms of discrete types have been proposed in three studies. They are based on rather different data: responses of teachers to a questionnaire regarding the reading habits of a sample of blind students (Lowenfeld et al., 1969), kimographic recordings made by attaching styluses to the fingers (Kusajima, 1974), video recordings (Davidson et al., 1980). Two pat terns are described in all three studies. In one, the left finger stays at the beginning of the line while the right one scans the text. In the other, which will here be called conjoint exploration, the two fingers scan the line and move to the next one side by side Other patterns involve dissociating the hands to various extents but here the studies diverge. Kusajima has two types: in one, the main part of the line is scanned by the two fingers side by side, but the left finger leaves near the end and moves to the next line while the right finishes the current one, in the other the left finger reads the first part of the line and the right the second part. Davidson et al. describe only one pattern of dissociation, where the first half of the line is read by the two fingers side by side and the second by the right finger alone. Lowenfeld et al's case consists of reading "next line with left hand before finishing preceding line with right hand". It is not possible to ascertain how serious the discrepancies between these studies are, since none gives quantitative definitions of category boundaries nor information on the crucial problem of subject consistency.

Dissociation of the hands is described with good quantitative details based on accelerated cinematography by Holland and Fertsch Eatman (1933) and by Fertsch (1947). These authors show that good readers dissociate the hands more than poorer ones.

Several often quoted assumptions regarding division of functions between the hands take no account of the variability of two-hand exploration patterns. For instance, Smith's (1929) proposal that only one finger collects textual information, the other one simply leading the way, or Heller's (1 904) notion that the right hand performs some kind of global pre-processing and leaves more detailed analyses to the left, make sense only for periods of conjoint exploration.

The present paper is devoted to a systematic analysis of the patterns of hand movements displayed by the blind readers who participated in the experiment described in the preceding paper (Mousty and Bertelson, this issue). The analysis will proceed in two stages. First, a general non-quantitative but systematic description of hand and finger activity will be given. It will be followed by quantitative analyses of special aspects that appeared to be of particular interest for understanding the functions performed by the hands.

Method

The recording situation, the reading material and the general procedure have been described in the preceding paper (Mousty and Bertelson, this issue). In brief, each of 24 blind readers read aloud, with each hand alone and with two hands, two pages of prose, one page of 7th-order approximation to French and one page of scrambled words, all printed in uncontracted braille. The hands were recorded by two television cameras, one taking the reading page from above, together with the display window of a digital timer, and the other the vertical position of the fingers above the page. A grid of vertical and horizontal lines traced on each page of text made the position of each braille cell clearly visible on the recordings. Each page contained 14 lines, each 40 cells long.

The recordings were analysed by playing them back on a monitor. In that situation, it was possible to ascertain the position of any finger on the matrix of braille cells for each frame. "Actograms" representing the lateral position of the reading finger(s) at selected time intervals, namely every 0.1 sec, could thus be constructed. This very tedious procedure could, however, not be resorted to systematically. About three dozen actograms, each representing 16 sec of reading, were constructed. They were useful for the choice of the particular aspects of hand activity to submit to more systematic analyses. These further analyses were based on the selection of some key events occurring during the exploration of a line, and the recording of position and time information for each of these events. Details of the procedure will be provided together with the results.

Results

Overall Description of Hand Activity in Reading

The description is based mainly on unaided observations made at the time of testing or later when playing back the recordings. Occasionally, the tape was stopped on a particular frame or advanced slowly by hand. Actograms were constructed for typical lines, and some are reproduced in Figures 1 and 2.

Figure 1

#FIGURE:1

Figure 2

#FIGURE:2

Although no particular finger usage was prescribed, our evidence suggests strongly that the subjects used only the index finger(s) for reading, whether with one or two hands. In the case of about half the subjects only the indexes were seen to touch the page, all the other fingers being kept in the air. In the others in whom other fingers touched the page, they mostly did so outside the line being read. In only one subject could the hypothesis of middle finger involvement not be eliminated through simple inspection of the recordings, but when that subject was later asked to read with the middle finger of her preferred hand, she could do so only very slowly. In all the coming analyses of exploration movements, it was assumed that the reading fingers were the indexes.

Three main categories of hand movements can be observed during text exploration, whether with one or with two hands: (1) forward scanning along the line, from left to right; (2) return movement of the hand to the next line; and (3) regression of the hand to a part of text that has already been explored.

Forward scanning is generally continuous, with little variation in speed. In some subjects, the finger is seen occasionally to stop on a particular character. On such occasions, it generally engages in sagittal up-and-down movements, suggesting that input of text information requires movement of the skin relative to the characters. One important outcome of our observations is that the reading finger has never been seen to lose contact with the page, for example jumping over some redundant word or passage: exploration of the text was in the present study always exhaustive.

Return movements are faster than forward scanning (see for example Figure 1 a, c and f). The finger is generally kept in contact with the page. When the target of the return movement is the beginning of the next line (i.e. in one-handed reading, or for the LH in two-handed reading) the finger sometimes misses it and lands either in the left margin or on the second or third character. In the latter case, scanning sometimes proceeds from that position, so that the earlier cells are skipped, as happens repeatedly in boxes (a) and (b) of Figure 1. In other cases, a fast regression to the first cell is observed (Figure 2 c). A similar phenomenon has been described for visual reading (McConkie, 1983). At the end of the line, the reading finger is often seen to overshoot, i.e. to proceed over one or two empty cells before initiating the return movement (Figure 1 c and e). Less frequently, one or two characters are skipped at the end of the line (Figure 2 c). This happens mainly when the last character is a punctuation mark.

Regressions are resorted to by all readers, but with wide individual differences in frequency. As for the other types of movements, the reading finger remains in contact with the page during regressions. Amplitudes range from one or two characters to several words. In some slow readers, regressions are so frequent that the distinction between progression and regression is somewhat blurred: exploration of the text appears in these readers to proceed through back-and-forth scanning (Figure 1 e).

In one-handed reading, the hand proceeds along the line from left to right, then returns to the beginning of the new line. There are no apparent differences between the operating modes of either hand.

In two-handed reading, two modes of hand cooperation can be observed, which we shall call respectively conjoint and disjoint exploration. In conjoint exploration, both indexes proceed along the line side by side, generally touching each other, or very nearly so, which means that the fingertips are about one or two cells apart. In disjoint exploration, the two hands, simultaneously or successively, explore different pas sages: while one hand explores one passage, the other hand returns to the line or explores a different passage.

In the vast majority of cases, the line is explored using a mixed pattern, which combines conjoint and disjoint exploration. Typically, the LH reads the beginning of the line alone, is then met by the RH returning from the previous line, and the two hands explore a median segment of text in the conjoint mode, until the LH leaves for the following line, letting the RH finish the current one. A typical example appears in Figure 1 b. Reading a line following the mixed pattern involves thus a left-hand segment (LHS), a two-hand segment (2HS) and a right-hand segment (RHS).

There are wide individual differences in the respective lengths of the three segments. In some subjects, the 2HS can reduce to one or two cells (Figure 2 c and d). Total elimination of the 2HS is however extremely rare, and was observed only for three of the 576 lines of prose we analysed. Apparently, the LH is not sent on its return journey until it has overlapped with the RH. At the other extreme, some subjects read the major part of the line in the conjoint mode. In these subjects, the LH nevertheless generally leaves the line a few cells before the end and performs its return while the RH finishes the line. An example of that kind of pattern appears in Figure 2 b. The LH, after reaching the beginning of the next line, either waits for the RH (as for the second transition of Figure 2 b), or starts scanning the new line (other transitions of the same box). Total conjoint reading, where the hands scan the whole line side by side and also return to the line together, is observed occasionally in some subjects (Figure 2 a), but in none of them all the time.

The subject who was eliminated from the main analysis because the performance of his LH was too inferior to that of the RH (Figure 1, Mousty and Bertelson, this issue) used an exploration pattern where the RH explored the main part of each line, while the LH was kept near the beginning, occasionally moving a few cells to the right (Figure 2 e). It seems likely that the LH played little role in picking up textual information and served mainly to help the reading RH localize the beginning of the line. Such use of the LH is nevertheless beneficial, for it allows a 29% gain in speed relative to reading with the superior RH alone. This exploration pattern, which as we have seen in the introduction figures in previous descriptions, could perhaps be called "assisted one-handed reading". One reason why it did not appear more frequently in the present study may be that readers who use it may generally not consider themselves as two-handed readers and so were eliminated by our recruitment procedure.

One interesting aspect of two-handed cooperation is that the LH is often seen to proceed along the line before the RH has completed the exploration of the previous line (Figure 2 c and d). In such cases, there is thus a period of time during which the hands explore simultaneously different passages of text. This operating mode will be called "simultaneous disjoint exploration" (SDE).

Patterns of Two-Handed Exploration:

Quantitative Analyses

Sizes of Segments Explored in the Different Modes

We saw in the preceding section that in two-handed reading the majority of lines are explored following a pattern that comprises three segments, LHS, 2HS and RHS, read, respectively, by the LH alone, the two hands and the RH alone. The length of these three segments was measured for the 12 median lines of each page read bimanually by each subject. The measurement was based on two key items of information per line: (I) the first cell reached by the RH reading finger on its return journey, and (2) the last cell reached by the LH reading finger before leaving the line. These two cells are the limits of the 2HS; LHS extends from the beginning of the line to 2HS and RHS from 2HS to the last character of the line. Empty cells traversed when the finger overshoots the end of the line are thus left out of the count. For the few cases (15 altogether) where a regression brought one finger back to a passage of text originally read by the other finger alone, the division into segments was always based on the initial exploration.

Given that % LHS, % 2HS and % RHS add up to 100, a diagram showing two of the three values contains all the useful information regarding the three variables, exactly as in a standard chromaticity diagram. Mean individual values of % LHS and % RHS for prose reading are given in scattergram form in Figure 3. The length of 2HS is represented by the position of each dot relative to the set of oblique 45 deg lines, with long 2HSs near the origin and short ones away from it.

Figure 3

#FIGURE:3

These data document the wide variability of individual exploration patterns mentioned in the preceding section. Mean 2HS ranges from 8.2 to 93.7%, and each mean one-handed segment from less than 5% to about 60%. Subjects vary also in the relative lengths of the two one- handed segments. However, a definite majority of subjects (19 out of 24) have a mean RHS longer than their LHS. By a simple sign-test, this deviation is significant at p=0.003.

Individual patterns are, on the other hand, very stable. Product-moment correlations between measurements made separately on the two pages of prose read by each subject reach 0.96 for 2HS and 0.94 for the ratio LHS/(LHS + RHS).

One important aspect of the results apparent in Figure 3 is that the distribution of individual patterns of hand usage is continuous rather than discrete: there is no obvious clustering that would invite a description in typological terms.

There is no systematic difference in pattern of exploration between the nine congenitally blind subjects and the other ones. For instance, mean % length of 2HS is 43.4 for the first and 45.7 for the latter.

To examine individual differences in operating mode further, subjects were classified according to size of 2HS into three equal groups.

Table I

#TABLE:I

Table I shows speed of one- and two-handed reading of prose for these three groups. All three reading speed scores (one-handed reading with, respectively, fast and slow hand and two-handed reading) go up strongly with decreasing 2HS. One-way ANOVAs of these scores, whose results are given in the table, yield a significant effect of subject groups for two-handed reading and for one-handed reading with the fast hand, not for the slow hand. Product-moment correlations between 2HS and the three speed scores, which are also shown, are all significantly negative. Two-handed gain tends to be larger in Group 1, but the difference falls short of significance, as does the correlation between two-handed gain and 2HS.

As already mentioned, individual differences in the relative sizes of the two one-handed segments show considerable reliability. A reasonable assumption was that these differences are related to the respective reading speeds of the hands as measured in one-handed reading. The product-moment correlation between the ratio LHS/(LHS + RHS) and relative superiority of the LH in one-handed reading (RLHS as defined in Table I, Mousty and Bertelson, this issue) is 0.75.

 Figure 4

#FIGURE:4

Figure 4 shows the relation of pattern of exploration to type of material. It appears that 2HS increases from prose to approximations and again from approximations to scrambled words. The effect is stronger in Group 1, weakest in Group 3. A 2-way (subjects x materials) ANOVA of 2HS yielded a significant effect of material [F(2, 42)=29.8; p<0.00l] and a significant Group x Material interaction [F(4, 42)=4.0; p<0.0l]. The effect of groups, which is of course significant, is spurious, as the groups were orginized on the basis of 2HS.

It seemed worthwhile to examine whether the point at which the LH starts its return movement is related to word boundaries. Toward that purpose, cells last explored by the LH on each line were classified into four categories: (1) inter-words interval, i.e. blank or punctuation mark; (2) first cell of a word; (3) last cell of a word; (4) "interior" of a word, i.e. all cells between the first and the last one. The obtained distribution is not significantly different from the one predicted on the assumption that word boundaries have no influence. This does not mean, of course, that the departure of the finger is not triggered by word identification: for the time being, we simply do not know at what moment, during or after exploration of a word, identification occurs.

Simultaneous Disjoint Exploration

One striking accompaniment of disjoint exploration, which has been mentioned already in the overall description, is that the LH often starts moving on the new line before the RH has reached the end of the preceding one. Four key items per line were used to measure the phenomenon: (1) the time at which the RH leaves the last character of the line, (2) the time at which the LH leaves the first cell of the next line, (3) position of LH reading finger at time (1), and (4) position of RH reading finger at time (2). Whenever (2) is earlier than (1), simultaneous disjoint exploration (SDE) is considered to have taken place, for a duration equal to interval (2)-(l). (3) and (4) make it possible to measure the number of characters explored simultaneously by the LH and the RH, respectively.

SDE is observed on the majority of prose lines. Four subjects only do not generally show it (i.e. on less than 3 lines out of 22). In 15 out of the remaining 20 subjects, it occurs on 15 lines or more. The mean numbers of cells explored simultaneously by each hand on those lines where SDE occurs are given in Figure 5 in seattergram form.

Figure 5

#FIGURE:5

The subject represented by the extreme right dot displays an exceptional pattern of movement and was left out of the analysis centered on SDE. For the remaining subjects, the mean number of cells read simultaneously ranges from 1.5 to 6 for the LH and from 1 to 8 for the RH. Apparent also on the figure is the fact that, as expected, fast subjects tend to perform more SDE than slow ones. Two-way ANOVA (subjects x hands) was applied to these data, and the main effect of subjects' speed [F(l, 17)=5.39; p=0.05] was significant while that of hand [F(l, 17)=4.18; p>0.05j and the speed x hand interaction were not.

For the interpretation of the SDE phenomenon, a critical question is whether it is accompanied by a slowing down of forward scanning movements: does each hand proceed along the line as fast when the other hand is also exploring as when it is exploring alone? A tentative answer can be given by calculating, separately for each hand, mean time per cell for the line segments explored during SDE (cases where regressions occurred during SDE were eliminated from the count), and comparing these values to time per cell during forward scanning in one-handed reading with the same hand. These computations are part of the time analysis described in the next section.

Table II

#TABLE:II

The results, which are given in Table II, separately for the 8 subjects with the larger and the 8 with the smaller number of cells read by the LH under SDE, are very clear. The subjects with little SDE reduce their forward scanning speed considerably during SDE, on the average by a factor larger than 2. On the contrary, subjects with superior SDE show little slowing down. Two-way ANOVA applied to the ratio (time per cell during SDE)/(time per cell in unimanual reading with the same hand) gave a significant effect of subjects' group (high vs. low incidence of SDE) [F(l, 14)=4.60; p=0.05] and no effect of hand and no group x hand interaction; t-tests applied to the differences between the two time scores in the high incidence group were significant for the LH [t(7)=3.13; p=0.0l] and fell just short of significance for the RH [t(7)=1.82; p>0.10]. These small differences do not necessarily reflect slowing down of scanning during simultaneous exploration, for scanning speed is not necessarily constant across the line, so that mean speed of scanning in one-handed reading is not the best baseline with which to compare speed during SDE. Examination of the actograms suggests in fact that in one-handed reading scanning speed can be slightly slower at the beginnings and ends of lines.

Table III

#Table:III

Table III shows total number of cells read under SDE per 100 words of text for prose and for scrambled words, respectively, for the high incidence subjects. Amount of SDE clearly decreases in scrambled words reading.

Analysis of Total Reading Time

Using the key events methodology, it is possible to analyse total time taken to read a passage into components devoted to different activities, namely (1) regressions, (2) line transitions, and (3) forward scanning.

Regression duration is the time elapsing between the moment one finger is seen moving backwards to a passage that has already been scanned, and the moment a new cell is reached in the new forward movement. The duration of a transition is the time for which scanning is interrupted between two lines. In one-handed reading, it is equal to the duration of the return movement of the reading finger. In two-handed reading, it is the time elapsing between the moment the RH leaves the last character of one line and that at which the LH begins scanning the new line. Whenever SDE occurs, transition time takes a negative value. Total forward scanning time is the total time spent scanning each segment of text. It can be obtained by default by subtracting the sum of regression durations and transition durations from total reading time. The convention to count duration of SDE as negative transition time implies that the corresponding duration is counted twice in total forward scanning time. In other words, for a subject who engages in much SDE, total forward scanning time can be longer than total reading time.

The three components were computed for each subject for the two pages of prose read with two hands and for those read with the fast hand, leaving as usual the first and the last line of each page out of the analysis. For transition times, positive and negative values were summed separately.

Table IV

#TABLE:IV

Mean total time and its analysis into the three components are given in Table IV for the three groups of subjects formed on the basis of length of mean 2HS. As could be predicted from the correlation of % 2HS with speed scores, total reading time increases from the more disjoint readers (Group 1) to the more conjoint ones, and relative increase is similar for one-handed and two-handed reading.

In one-handed reading, the relative weights of the three components do not change much across the three groups. The main difference is that regression time increases from Group 1 (12.5%) to Group 3 (19.0%). Forward scanning time represents 70% of total reading time for Group 1 and 66% for Group 3. The gain from two-handed reading is in all three groups due mainly to the reduction in transition time. The major reduction results from the elimination of scanning interruptions. In Groups 1 and 2, a further gain results from SDE. In these two groups, the reduction in overall transition time accounts for the whole of the two-handed gain. In Group 3, in which very little SDE is observed, additional gains come from reductions in regression time and in forward scanning time

Discussion

When reading with two hands, most of the present subjects adopted mixed movement patterns combining two principles of hand cooperation. We called these principles conjoint exploration, which consists of the two reading fingers proceeding side by side along the same passage, and disjoint exploration, where the two fingers scan different passages. For the vast majority of lines, a first segment is scanned by the left hand alone, a second one by the two hands conjointly, and a last one by the right hand alone. Huge and highly reliable individual differences have been observed in the respective sizes of the three segments.

The lengths of the different segments vary in a continuous way, without the sort of clustering that would signal the existence of discrete types. Cases of nearly totally conjoint reading or those with nearly total dissociation of the hands are more parsimoniously considered as end points on a continuum of mixed exploration patterns than as special categories.

We saw in the introduction that exploration patterns have been described in terms of discrete types in several studies. In none of these, however, was the kind of evidence provided that would be necessary to establish a typology. In Lowenfeld et al.'s questionnaire, the categories into which the respondents had to classify the operating modes of their students were provided by the authors, without quantitative criteria. The existence of types was thus taken for granted. The other two studies (Kusajima, 1974; Davidson et al., 1980) are based on actual observations of hand movements, but there are again no quantitative descriptions of category boundaries and no information regarding subject consistency (except for Davidson et al.'s statement that their subjects were "extremely consistent in using only one of the four global styles").

The recourse to disjoint exploration is strongly correlated with speed of reading, a result that confirms those of Holland and Fertsch Eatman (1933) and Fertsch (1947).

This correlation suggests strongly that conjoint exploration might represent a less advanced operating mode than disjoint exploration and that readers might with increasing practice shift from the first to the second. In many schools, pupils are first taught to read in the conjoint mode, and discover later, with more or less encouragement from the teacher, that they can dissociate the hands. In Lowenfeld et al.'s survey, dissociation of the hands was reported more frequently for grade-8 than for grade-4 pupils. This important notion of a progressive shift from conjoint to disjoint exploration should be validated through actual longitudinal studies.

The analysis of exploration movements has revealed two main rea sons why dissociating the hands might allow faster reading. One is reduction or elimination of interruptions of scanning by having each hand perform its return movement while the other is reading. The second reason is what we have called simultaneous disjoint exploration-simultaneous scanning of the end of one line by the right hand and of the beginning of the following line by the left hand. In good disjoint readers, we have shown that simultaneous exploration can be achieved with very little reduction of speed of forward scanning, which allows twice as much textual information to be collected per time unit as during one-handed reading.

For the two preceding factors to operate, a limited degree of hand dissociation is sufficient. They thus provide no explanation for the fact that some readers dissociate the hands nearly completely. Another factor that might be worth considering is that in conjoint exploration each hand performs a larger amount of movement over the page-i.e. more forward scanning and also larger return journeys-than in disjoint exploration. This factor is probably not very important during short spells of reading, but it might become so in the long run. The larger amount of movement involved in conjoint reading might cause more muscle fatigue. On the other hand, the numbing of tactile sensitivity which according to some braille readers occurs after prolonged reading spells (Javal, 1904) might be delayed under disjoint exploration.

At first, one would expect the correlation between speed of reading and hand dissociation to be due to higher efficiency of disjoint exploration per se. Several other aspects of our data show that this interpretation would be an oversimplification. First, although the correlation is highest with speed of two-handed reading, it exists also with speed of one-handed reading. More disjoint readers tend thus to be faster one- handed readers. The analysis of total reading time has further shown that only one third of the superiority of the more disjoint readers is obtained through faster line transitions, and that the rest results from faster forward scanning and less time spent on regressions.

The superiority of disjoint reading does in consequence not reduce to a one-way effect of the adoption of a more economical operating mode. Hand dissociation appears rather as one of several sub-skills that are acquired with reading experience, another one being the ability to scan the text fast and reliably with each hand separately. This conclusion receives support from the study of Wormsley (1981) who tried to train pupils dissociating the hands, and found that this training did not affect their two-handed reading style outside the experimental situation.

Conjoint exploration must also present some advantage, otherwise it would be difficult to understand why most readers use it on some part of each line. Two of our findings further point to the existence of an advantage. The analysis of total reading time has shown that in the most conjoint readers forward scanning is faster in two-handed than in one- handed reading. It is tempting to assume that the gain is obtained specifically during the two-handed segment. To test the assumption, one would need to measure directly scanning speed during the different segments. Such measurements would have been prohibitively time- consuming with the present methods, but will become possible with a forthcoming recording system. An alternative solution is to have subjects read comparable passages with one hand and with two conjointly. The other relevant finding, that readers increase the length of their two-handed segment when reading materials with lesser contextual constraints, suggests strongly that conjoint exploration yields more reliable textual information than disjoint exploration. This factor might be invoked also to explain the relation between speed of forward scanning and hand dissociation: with increasing ability to collect text information with each finger separately would come a reduced need for the help from a second finger.

The respective resort to the hands is correlated with the speed they attain in one-handed reading, at the level of which exist large and reliable individual differences. Two kinds of effects may have brought about that correlation. First, readers may discover by trial and error the pattern of hand dissociation that makes full use of the reading capacities of the two hands. That pattern must be one where both hands take the same time scanning their allotted segment and performing their return journey. The other kind of effect is that a pattern of dissociation that leaves a longer segment to one hand will provide that hand with more practice opportunity and improve its reading ability. Only longitudinal studies of the development of exploration patterns could bring information regarding the respective roles of the two kinds of influences.

The preceding notion of an interaction leading to mutual adjustment of individual hand reading speed and pattern of hand dissociation does on the other hand not account for the fact that a majority of subjects used the right hand more than the left one in two-handed reading. Some additional factor, linked specifically to the two-handed reading situation, must operate to produce that asymmetry that has no counterpart in speed of one-handed reading.

Deeper understanding of hand cooperation requires more systematic analyses of the mechanisms of two aspects of hand activity: simultaneous disjoint exploration and conjoint exploration.

1. Simultaneous disjoint exploration had been mentioned before (Grasemann, 1917; Holland and Fertsch Eatman, 1933; Lowenfeld et al., 1969) but had not been analysed in any detail. We have shown that in the subjects who resort strongly to this exploration mode, little slowing down of forward scanning is observed during the period of simultaneous exploration relative to the speed measured during one-handed reading. In those readers, textual information is thus being collected by the two hands in parallel. The finding strongly suggests the operation of some buffer store, where the left-hand data are preserved until they can be integrated with those picked up simultaneously by the right hand.

The notion of a tactile sensory store has already been examined by Bliss, Crane, Mansfield and Townsend (1966) using Sperling's partial report procedure. The subjects received patterns of local stimulations on the fingers of both hands. In the total report condition, the subjects tried to recall the stimulus locations for the two hands, while in the partial report condition the recall concerned only one hand, indicated after stimulus cessation. A small superiority of partial report was obtained, which suggested a buffer store with a very small capacity, of the order of one point of stimulation. However, the stimulation patterns used in that study, spatial configurations of simultaneously stimulated points, are very different from the dynamic sequence of reafferent stimulations provided by active exploration of braille and, despite the extensive training received by the subjects, had not been learned to the extent that braille symbols are in skilled readers. It is clearly necessary to apply the partial report procedure to blind subjects with braille material.

2. We have obtained rather good confirmation that conjoint reading with two indexes side by side allows more reliable collection of text information. This aspect of performance also calls for further study. The advantage might be due to some division of functions between the fingers, but not necessarily. A parsimonious hypothesis would be that joining the fingers is helpful simply because it enlarges the window of text from which tactile information is being gathered simultaneously.

REFERENCES

Bertelson, P. and Mousty, P. (1982). Modes opératoires dans la lecture de l'écriture braille. In J. Leplat and P. Bertelson (Eds.), Hommage ä J. M. Faverge. Le Travail Humain, 45, 13-23.

Bliss, J. C., Crane, H. D., Mansfield, P. K. and Townsend, J. T. (1966). Information available in brief tactile presentations. Perception and Psychophysics, 1, 273-283.

Bürklen, K. (1917). Das Tastlesen der Blinden-Punktschrift. Beiheft zur Zeitschrift für angewandte Psychologie, 16, 1-66.

Davidson, P. W., Wiles-Kettenmann, M., Haber, R. N. and Appelle, S. (1980). Relationship between hand movements, reading competence and passage difficulty in braille reading. Neuropsychologia, 18, 629-635.

Fertsch, P. (1947). Hand dominance in reading braille. American Journal of Psychology, 60, 335-349.

Foulke, E. (1964). Transfer of a complex perceptual skill. Perceptual and Motor Skills, 18, 733-740.

Foulke, E. (1982). Reading braille. In W. Schiff and E. Foulke (Eds.), Tactual Perception. Cambridge: Cambridge University Press, 168-208.

Grasemann, P. (1917). Eine Untersuchung über das Lesen der Blinden. Beiheft Zur Zeitschrift für angewandte Psychologie, 16, 67-72.

Heller, T. (1904). Studien zur Blindenpsychologie. Leipzig: Engelmann.

Holland, B. F. and Fertsch Eatman, P. (1933). The silent reading habits of blind children. Teachers Forum, 6, 4-11.

Javal, E. (1904). Der Blinde und seine Welt. Hamburg-Leipzig: Voss.

Kusajima, T. (1974). Visual reading and braille reading: An experimental investigation of the physiology and psychology of visual and tactual reading. New York: American Foundation for the Blind.

Lowenfeld, B., Abel, G. L. and Hatlen, P. L. (1969). Blind children learn to read. Springfield, Ill.: Thomas.

McConkie, G. W. (1983). Eye movements and perception during reading. In K. Rayner (Ed.), Eye movements in reading: Perceptual and language processes. New York: Academic Press.

Mousty, P. and Bertelson, P. (this issue). A study of braille reading: 1. Reading speed as a function of hand usage and context.

Smith, J. M. (1929). Which hand is the eye of the blind? Genetic Psychology Monographs, 5, 209-252.

Wormsley, D. P. (1981). Hand movement training in braille reading. Visual Impairment & Blindness, 75, 327-331.

Manuscript received 21 November 1984

Requests for reprints should be addressed to Paul Bertelson, Laboratoire de Psychologie expérimentale, 117 Av. Adolphe Buyl, 1050 Brussels, Belgium.

This work was partially subsidized by the Belgian "Fonds de la Recherche fondamentale collective" (F.R.F.C.) under Convention 2.4505.80, by the Research Council of this University and by the Ministry of Scientific Policy (action de Recherche Concertée "Processus cognitifs dans la Lecture"). The cooperation of the readers who volunteered their services during long recording sessions is gratefully acknowledged. Thanks are due to Mr Claude Sehepens of the "Ligue Braille" and the Directors of the institutes for the blind of Berchem-Ste-Agathe and Woluwé-St-Lambert for their help. Barbara Friz and Viviane Kurrels assisted in analysing the data.