Text-scanning patterns of blind readers using Optacon and braille

VeteransAdministration

Journal of Rehabilitation Researchand Development Vol . 22 No . 3BPR 10-42 Pages 54-65

Text-scanning patterns of blind readersusing Optacon and braille

DAVID W. HISLOP, PH .D.*; P . L . ZUBER, PH.D.*, AND JOHN L. TRIMBLE, PH .D.Rehabilitation Research and Development Centel, Veterans Administration Hospital,Hines, Illinois ; and *Department of Bioengineering, University of Illinoisat Chicago, Chicago, Illinois 60680

Abstract—Tactual reading usually requires the coor-dinated use of the forearm/hand motor and tactilesensory systems. Therefore, in a series of studies oftactual reading behavior, we chose to record the text-scanning patterns of adult blind readers in an attemptto gain further insight into the nature of this process.Reading behavior for four modes of tactual readingwas recorded (embossed braille one-hand, embossedbraille two-hands, Optacon/letterprint, and Optacon/Inkbraille) by detecting and recording the instantane-ous position of light-emitting diodes unobtrusivelyattached to the finger(s) or the Optacon camera.

In addition to the comparative evaluation of the foursets of reading patterns, the salient features of theOptacon/letterprint patterns were quantitatively ana-lyzed in an attempt to characterize this particularmode of reading . The text-scanning patterns of Opta-con readers have not been previously reported.

In general the text-scanning behavior for all modesof tactual reading seems to be similar; the onlyremarkable difference appears to be in the readingrates. Regarding Optacon/letterprint performance,reading rate was found to be significantly and nega-tively correlated with line-changing time and thenumber of regressions . No significant correlation wasevident between rate and regression magnitude.

INTRODUCTION

Text-scanning effected with an appropriatemotor system is a concomitant of the reading

This work was supported by the VA Rehab . R&D Service,and a Biomedical Research Support Grant from the Univer-sity of Illinois at Chicago.

Address for reprints : B .L. Zuber, Department of Bio-engineering, University of Illinois, Chicago, IL 60680 .

task. Sighted reading requires the coordinateduse of the oculomotor and visual sensory sys-tems; similarly, tactual reading by blind individu-als entails the coordinated use of a motor and asensory system. Specifically, tactual reading usu-ally involves the use of the forearm/hand motorand tactile sensory systems.

Tactual reading by blind individuals is accom-plished primarily with embossed braille (9) . Asecond system of tactual reading, which is cur-rently used to a limited degree, involves anelectronic reading aid called the Optacon thattransforms common letterprint materials into atactually readable form (3) . Tactual readingeffected with either embossed braille or theOptacon system is slow compared with sightedreading. Tactual reading rates range from aboutone-tenth (Optacon/letterprint) to about one-third (embossed braille) those of sighted readingrates (8, 10).

Presently, the complex task of reading is poorlyunderstood . Consequently, a cogent explanationof this process is virtually impossible to formu-late . This lack of a fundamental understanding isquite apparent when an attempt is made tospecify the basis for the great difference betweensighted and tactual reading performance . Also asdisappointing, there is no meaningful explanationof the difference between embossed braille andOptacon/letterprint performance . Fortunately,the text-scanning movements of readers are rela-tively simple to monitor and record ; furthermore,

54

55

HISLOP ET AL . : Optacon/Braille scanning patterns

,,i7 7 r . "0. . 3 ° '

:I 'F 3 7 :°3 -3 73 : ;

7 3 3 3Li 3 3 " E 73 :

r

s:

.7 3 asl a rr: 33 : : 31 3 .3

3 1 i-in

.3h 747 °,4:

2r7 :7 h e q L arlis' XI Pr°. re rl :. 3 a

is° °l a

E

C a sss

rh 3r e 3isss a E ri a "i 3C

.#'F'3

. .3 3 :37s :7if. li E

sl a in

.5 #33a

Ce r ts3 sl 3iii -1 3 3 3 "s 3ra ss is: r

Nis i

3r a 1'7 ar &:3

7 13

#33ian

.s :°

3"hii :' :,

La:

Is3 ii a ° sr i :2 . ts:

737 s %

7 3 i rsr ss 'ehf ss a rlsi-* al° r r°. :° 3 h 3i r :4 3 .7s:

P sla'3 E #3 1 3 1s r :7 ss

.'rs

s:7 33 73 4 " 3 a 3 s: ss

3°ii

.Ea°l

"I %

P :13 1a s sss:s

L :sr ' 3 ss°"2a

3 ' ss -sq

a3 as' E Lals.'saas3 ii- 3 3irs: aan

:13 i

si as ls o°3

s .13r a

rl7 ss ss ss r ah s ae s:

h 37 .7 37 •

" . as' 30ss 'sna ara

4 3%rs' ss '3 rss i-3 :4 133 : era . .i3 r

33:.a r

"•ss .®#03 7''#- rlsfs® %3" : a 73 :4 9'4 r i,

.i"h33 .aiL as-3 3h

o1- 3 "31sl ias°

. s i

Is3irn mi s s

31 :1 3 1 a ia L :s h ar :3

r7 1 : 3

3 3ri ssr 'ss 's # s3 a e

7'#n

sir '3 3h r '3 a 5s, LI (

73:

I ss r ®'

s sa se *ss ss ar r q r ssllr

r s e

ss

ass% ( 3 1 3 ' ss h 3 :4 3 .:s 3LT 1 14 13— :a is :

r is ssi ra3ls ''i

ii3is * 77a

h"L•s

sr q s''37 1 f is sri,- :°i

rh3 r h°7

"Oa :a '

r• " 0®•1°

L ." 3 3 — as°°tss # 3 h 3r :th l i 337 ss

! 1.33 :s 3 'r3 r7 s : 3

r 3 s° 3 s: :Es" .7i ra °i

#

a ss 3a 3

3 3 3 73 "Oa

Ls a Ii3ss a h T a r"'

E ai' : 3 s' L ana i : 3 n

4 7 : ''I

'if,

a l 3 1 •:7 1s°s.

:: ,33s° 3aL .3

h 3r 3h 3 ar . E 3 37 :l e a 7 a7 h3r :47a a 3 %7 a P3 3 7 .

a 331

FIGURE 1Sample of Inkbraille (a reduced-size, ink-image version of brain

ill . Sample is somewhat distorted by reproductionfamiliar braille code) . Inkbraille is read electronicall

however, each dot of an Inkbraille text is typicallyan Inkbrailler whose output is a "standard-size"

id distinct .

147 3

i

56

ournal o

litation R s r- :--I and Development Vol . 22 No . 3 July 1985

nd

importantly, analysis of this overteased manifestation of informa-

acquisition behavior can reasonablyted to lead to greater insights regarding

iracter and nature of the overall process.:lie temporal forearmlhand movement patterns

are a precise means by which to quantify bothmean and instantaneous (forward prog : scionalong a line of text) reading rates . These pa nscan also be used to study the frequenc y andextent to which readers choose to retrace por-tions of text they have already scanned. Whenobtained under controlled conditions, the ob-served frequency, extent, and rate of rescanningmovements could help elucidate the manner inwhich tactile readers handle semantic informa-tion acquisition. In addition, these patterns per-mit the precise segmentation of observed readingperformance into a text information–acquisitionphase and a line-changing phase . Given theirpotential usefulness, the text scannin ,–trackingpatterns of tactual readers and a comparison ofthe patterns generated during several modes oftactual reading were chosen as the focus of thisstudy.

This report presents the text-scanning pat-terns of experienced adult tactual readers usingthe embossed braille (one-hand and two-handreading) and Optacon/letterprint modes of read-

ing. In addition, a sample of the performance ofreaders using the Optacon to read a reduced-sizeink-image version of the braille code (Inkbraille)has been included for comparison . These data aretypical examples of the performance of as manyas 20 practiced adult blind readers, some of whomparticipated in as many as five experimentalsessions.

Embossed BrailleEmbossed braille has a rich history of about

150 years (13) during which time the code and itsformat have been standardized . The braille codeis composed of code units called cells . Each cellmay contain up to six , raised-dot elements ar-ranged in two vertical columns of three dots each(alternatively, three rows of two dots each) . Thebraille reader is expected to discriminate be-tween the dot patterns (63 possible) formed byfrom one to six dots . The basic braille code (gradeI) involves mostly a one-for-one encoding fromletterprint to braille-encoded text : each braillecell is used to represent a letter, numeral, orpunctuation mark . However, other versions ofthe braille code (e .g., grades II and III) use"contracted forms" to represent with fewer braillecells common letter combinations and frequentlyused words . This description of braille is neces -sarily brief for the purpose of this discussion;

FIGURs 2Horizontal (top trace) and vertical (bottom trace) hand positions recorded during embossed braille reading constrained to use ofone hand . Traces represent reading of four complete lines of embossed braille . Reading of each line is followed by a rapid right-to-left line-changing movement . Regressions are indicated by smaller right-to-left movements .

97

H SLOP ET AL . : Optacon .Braille scam-

and

Horizontal Poaitio

Lint,

d

At

Horizontal Position

Vertical Position

FIGURE 3Hand position recorded during "normal" embossed braille reading . Upper two traces are position (horizontal and vertic .fright hand during reading, and bottom two traces are position (horizontal and vertical) of left hand . This reader used left ha]line marker only.

however, those desiring a more detailed descrip-tion of embossed braille are referred to thearticle by Foulke (8).

Optacon System

The Optacon tactile reading system is a recenttechnological advance, which was made commer-cially available in 1971 (17) . Reading with theOptacon involves scanning text with a smalldevice (electronic camera) held in one hand andinterpreting a vibrating-pin tactual facsimile ofthe text with a finger of the second hand . Theportion of text imaged by the electronic camera isapproximately one letterspace . Functionally, a 6-column by 24-row monolithic array of photosens-ing elements is coupled with a matching piezo-electrically driven array of 144 pins . The tactualdisplay pins, which are activated at any instant,

are those whose paired photosensing elhave a dark field imaged onto the : sensinConsequently, as text is scanned U'outpuOptacon system is a tactual "Tim; Simarldisplay. Conceptually, the Oldesigned primarily, and al):priately, for reading lett~however, to transduce (contrast. An err-amp'Optacon to read rn

Inkbraille

Possible basis12), A.

sh

figure 1 . Ithen Inkbraille is read with an

led reading aid (Inkbrailler),is approaching those of em-

Inkbraillforsamlis hoped thatappropriately

readir

58

Journal of Rehabilitation Research and Development Vol . 22 No . 3 July 1985

bossed braille will be attainable . In the absence of METHODSan Inkbrailler a conveniently available Optaconwas used to test the readability of Inkbraille.Given this opportunity to test Inkbraille and atthe same time observe the effects of its use onOptacon reading performance, we recorded theOptacon/Inkbraille reading performance of someof our readers.

Previous Studies of TactualReading Performance

Purklen (5) and Kusajima (14) have previouslyreported on studies of directly recorded em-bossed braille reading patterns . Purklen attach-ed pencils to the index fingers of braille readers;Kusajima used a similar method to producetracings on the surface of a paper-covered drumwhich was rotated at a constant speed . In princi-ple, Kusajima's results are analogous to thosepresented here . Other investigators such as Eat-man (7), Davidson et al. (6), and Bertelson andMousty (2) used indirect methods (motion pic-tures and videotapes) to acquire a record ofreading performance . Although embossed braillereading performance has received considerableattention, at least in the areas of comparativeperformance of proficient and poor readers andthe idiosyncratic manner in which two hands areused for reading, Optacon reading performancehas received little attention. Moreover, compari-sons between embossed braille and Optacon read-ing—text scanning patterns have not previouslybeen made.

The paucity of Optacon/letterprint reading per-formance data prompted our attempt, as part ofan ongoing series of tactual reading experiments,to develop baseline measures of` this behavior.Consequently, more than 800 lines of Optacon/letterprint silent-reading performance were re-corded and analyzed . Three of the salient fea-tures of these text scanning—tracking patternswere found to be 1) line-changing time, 2) num-ber of regressions (right-to-left movements dur-ing the reading of a line of text), and 3) mag-nitude of the regressions. The influence of thesefeatures of tactual reading performance on read-ing rate were examined and are reported herein .

The collective performance of 20 adult blindproficient readers was recorded . Some of thesubjects participated in as many as five experi-mental sessions in which they read embossedbraille texts (with one hand and two hands) andused the Optacon to read letterprint and Ink-braille texts . A number of subjects read as manyas 40 texts, whereas others read only 10 textsduring a series of studies . Individual subjectswere selected for the separate studies based onthe practiced reading skill(s) required for eachexperimental task.

The 20 subjects used in these studies weremales and females ranging in age from earlytwenties to one subject who was in her middlesixties, as well as a number of subjects over 30years old . All subjects were college graduateswith residual vision limited to no more than lightperception . Except for one subject who was deafas well as blind, none of the other subjects hadany apparent limitations other than loss of vision.

Embossed braille texts were generally com-posed of lines with 40 braille cells. Inkbrailletexts consisted of lines averaging 50 cells inlength. The letterprint texts were composed oflines averaging 70 letterspaces in length . Virtual-ly all texts were of "block" format and consistedof 8 to 10 lines. Most of the texts were selectedfrom a Reader's Digest column called "Life inThese United States ;" however, of the approxi-mately 75 texts that were used, 35 were from aset previously rank-ordered in accordance withthe level of text difficulty (15).

The embossed braille texts were produced ingrade II braille code using a Perkins brailler.Single-spaced letterprint texts were producedwith a standard typewriter equipped with aPrestige-Elite 96 type font (12 letterspaces perinch) . Inkbraille (grade II) was produced usingan Apple II computer, appropriate software, anda Diablo daisy-wheel printer. The version ofInkbraille used for the pertinent studies allowed298 Inkbraille cells to be printed in 10 squareinches of text space (compared with 100 em-bossed braille cells per 10 square inches).

During the experiments the subjects wereasked to read silently and in a normal fashion .

59


FIGURE 4Horizontal (top trace) and vertical (bottom trace) hand-camera position recorded during letterprint reading with Optacon.Scanning for reading of three lines . Regressions show backward (right-to-left) hand movement . Horizontal hand velocity isshown in middle trace . Velocity channel saturates for line-changing movements and large regressions . Note difference in timescale between this figure and Figs . 2 and 3.

Furthermore, to help ensure normal readingbehavior the subjects were advised before eachexperiment that they would be tested on thematerials they had read . After completing thereading of each text the subjects were asked tosummarize the text and/or select from a word listthe word or words (if any) that had appeared inthe material just read. Regardless of the tactualreading modality, all experimental subjects ap-proached verbatim recall of text content.

Tactile reading performance was recorded asthe subjects read silently while seated comfort-ably at a reading table in a quiet laboratory, . Thereading surface of the table was monitored fromabove using a commercially available four-chan-nel Hamamatsu position-sensing system . Thebasis for recording the text-scanning be]of the experimental subjects was the dt e :t

of the instantaneous position of light- idiodes (LED's) within the reading field.LED's were attached unobtrusively to the indcfinger(s) of the readers hand(s) for braille p

or to the Optacon camera for letterprint orInkbraille reading . Voltages proportional to thehorizontal and vertical position of the LED's weregenerated by the instrumentation, and thesevoltages were recorded . In the case of theOptacon/letterprint study, the horizontal positionsignal was differentiated to produce a horizontalscanning velocity signal, and this was also record-ed. All data were recorded on a four-channel stripchart recorder. Spatial resolution was better thanone text character.

All four types (one-hand embossed braille, two-hand embossed braille, Optacon/letterprint, andOptacon/Inkbraille) of recorded text-scanningpatterns were qualitatively compared . In addi-tion t bdth the qualitative evaluation and theei

N, ith text-scanning patterns of otherreading investigated, the Optacon/

i 't reading patterns were subjected toquantitative evaluation . More than 800

tt ,anning traces for 10 readers, each reading10 texts of 8 to 10 lines each, were analyzed .

60


300

275

45 .8 4.

FIGURE 5Reading rate (Optacon/letterprint) as letterspaces perminute (LSPM) and "adjustedwords per 'minute" (AWPM) ver-sus number of regressions per100 letterspaces (R/100LS).Each data point represents meanperformance of a single subjectreading 10 texts .

37.8 —

1 I I6

9

12NUMBER OF REGRESSIONS/100 LS

Three salient scanning/tracking features of thesetraces were defined, and their influence on read-ing rate performance was examined. The mostprominent feature of the traces, which has beenreduced to a measure of scanning/tracking per-formance, is the mean time taken (by eachreader) to change from one line of text to the next(i .e., line-changing time). Another rather fre-quently occurring feature of the traces is theresult of small right-to-left movements (calledregressions) made by the readers as they read aline of text . Two measures of scanning/trackingperformance were derived from the regressionmovements . These two measures are 1) the mean

number of regressions per 100 letterspaces re-corded for each reader and 2) the mean mag-nitude of these regressions measured in letter-spaces.

Although embossed braille and Optacon/Ink-braille reading performance is reported here asthe familiar "words per minute," Optacon/letter-print rate performance is reported differently foranalytical reasons . Several Optacon/letterprintreading-rate measures were used in the analysisof the recorded data ; however, the two measuresof rate that best serve the purpose of this workare 1) letterspaces per minute and 2) "adjustedwords per minute."

HISLOP ET AL . : Optacon'Brai" .

Letterspaces per minute, rather than wordsper minute, is the preferred rate measure foranalysis of reading performance data . The reasonfor this is that reading rate measured in letter-spaces per minute is virtually independent of thetext from which the rate measure is derived (11).It has been previously reported that reading ratefor both sighted (15, 20) and tactual (11) readingis constant for a given individual regardless of thelevel of text difficulty, when rate is based on unitssmaller than words . Rate measured as words perminute is influenced by text difficulty due to textcharacteristics unrelated to reading behavior.'I\vo texts of the same physical length will be readin the same amount of time . Roth texts containthe same number of letterspaces ; therefore, therate in letterspaces per minute will be the same.If one text is more difficult, it will contain longer(and fewer) words. When rate is calculated inwords per minute it will appear that the moredifficult text is read at a lower rate, simplybecause this text contains fewer words . Usingrate measured as letterspaces per minute shouldcontribute to a more reliable interpretation ofreading behavior data . However, words-per-minute rates can conveniently be used for analyt-ical purposes if they are "adjusted ." Since wordsper minute is the most traditional measure ofreading rate (and when adjusted the resultspermit the same level of generalization as doesrate computed as letterspaces per minute) thereading-rate data of the Optacon/letterprintreading studies are reported both ways . Accord-ing to one source (1) five letters per word is agood estimate of the mean length of Englishlanguage words. Therefore, it is reasonable toconsider an average word plus associatedwordspace as being six letterspaces in length.Adjustment of the words-per-minute rate datawas therefore accomplished by dividing by sixthe reading rates, calculated as letterspaces perminute.

RESULTS

In our studies a mean embossed-braille readingrate of 116.1 words per minute was attained by agroup of five subjects (each reading five texts of

grade II braille), when readingread (unconstrained opportunit,hands) . This same group of five su106 .8 words per minute (each reading fiwhen the reading process was limited tof one hand. This observed level of perforis well within the range of rates pr_reported in the literature (8) . As regardscon/letterprint performance, a mean Optact,letterprint reading rate of 25 .2 adjusted wordsper minute (151 letterspaces per minute) wasattained by one group of subjects (10 subjects/10texts each) . Another group of 5 subjects (15 textseach) attained an Optacon/letterprint rate of 22.7adjusted words per minute (136 letterspaces perminute) . This observed Optacon/letterprint read-ing performance is similar to previously reportedOptacon reading rates (19) . Optacon/Inkbraillereading performance was determined to be 29 .5words per minute for a group of 5 subjects (eachreading three texts of grade II Inkbraille) . TheseOptacon/Inkbraille results are unique in thatthere are no comparisons available.

Figures 2 and 3 are samples of text-scanningpatterns recorded during embossed braille read-ing. Figure 2 is a sample of the data generated byone-hand braille reading . Figure 3 shows pat-terns for two-hand braille reading . Note that the(left-to-right) forward movement associated withreading a line is a ramplike trace of relativelyconstant average slope. This is indicative of arelatively constant rate of reading. The moresteeply inclined segments of the trace, with aslope opposite to the forward reading move-ments, reflect the rapid movements from the endof one line to the beginning of the next line . Thisline-changing action by the reader can be con-firmed by an inspection of the vertical positiontrace. Another salient feature of these records isthe occasional small, (right-to-left) motions indic -ative of a regression movement during reading.

Figure 4 is a sample of Optacon/letterprintreading performance . (Please note that the timescale in Fig . 4 is different from that of Figs . 2 and3 .) It is readily observed that, except for thedifference in slope (average slope of a line-reading trace could be used to estimate readingrate), these Optacon/letterprint reading tracesare very similar to those associated with em-

62

Journal of Rehabilitation Research and Development Vol . 22 No. 3 July 1985

FIGURE 6Net reading rate (Optacon/letterprint)versus line-changing time . Net readingrate (calculated by deducting line-changing times from total readingtime) is shown as letterspaces per min-ute (LSPM) and "adjusted words perminute" (AWPM) . Each data point rep-resents a single subject reading 10texts .

LINE CHANGING TIME (SECONDS)

bossed braille reading. Analysis of more than 800line-reading traces of Optacon/letterprint readingperformance resulted in these findings : Overallmean line-changing time (10 subjects) was 3 .0seconds, with a range of 1 .2 to 4 .5 seconds and astandard deviation of 1 .4 seconds . Overall mean(10 subjects) regressions per 100 letterspaceswas 7.4, with a range of 1 .7 to 14.6 and astandard deviation of 4 .6. Overall mean regres-

agnitude (10 subjects) was 4.4 letterspacesi range of 3.1 to 5.9 and a standardon of 1 .2 .

ter rint reading, the threeires were treated as

independent variables and reading rate was des-ignated as the dependent variable . Linear corre-lation analysis was used to identify trends . Anal-ysis of the rate versus regression magnitude datafailed to uncover any linear correlation (5 percentlevel) between these two variables . Figure 5 is aplot of rate versus the calculated mean number ofregressions per 100 letterspaces; each data pointrepresents the mean performance for a singlereader reading 10 texts . Analysis of the data inFigure 5 resulted in the finding that reading rateand the regressions per 100 letterspaces aresignificantly and negatively correlated (r = 0 .669,5 percent level) .

63

HISLOP ET AL : Optacon/Braille scanning patterns

Horizontal Position

Line I

Vertical Position

FIGURE 7Horizontal (top trace) and vertical (bottom trace) hand-camera position recorded during Inkbraille reading with Optacon . 'II-acesrepresent reading of three complete lines of Inkbraille . Time scale is the same as in Fig . 4 . These patterns are similar to thosefools embossed braille (Figs . 2 and 3) and Optacon/letterprint (Fig . 4).

Figure 6 is a plot of reading rate versus line-changing time. In this particular plot the readingrate can be described as "net reading rate" sincethe time required to change lines has beendeducted from the total reading time in order toeliminate the direct influence of line-changingtime on the reading rate . Each data point is thecalculated mean performance for a single subjectreading approximately 80 lines of text . With theuse of the linear least squares regression analysisprocedure, a significant and negative correlation(r=0.885, 5 percent level) was found between netrate and line-changing time . In 7 out of the 10cases, the individual subject data (rate) were notcorrelated with line-changing time . Rather, line-changing time appeared to be relatively constantfor each subject.

Figure 7 is a sample of the Optacon/Inkbraillereading performance that was recorded . [Pleasenote that the time scale in Fig . 7 is the same as inFig. 4 (Optacon/letterprint reading patterns) .]These tactual reading patterns are similar inform to those for the other two tactual readingmodalities . As with the other patterns, the sa-lient features of these traces are 1) ramplikeforward reading movements from left-to-right, 2)rapid right-to-left line-changing movements, and3) occasional moderate-velocity, small right-to-left regressive movements.

DISCUSSION

Apart from the difference in the slope of theforward-movement traces (left-to-right) in thecourse of reading a line, there appears to be noremarkable difference between the patterns re-corded for the several modes of tactual readingthat we have investigated . The slope of the line-reading traces (and the associated time scale forthe recordings) is indicative of the observed ratesof reading for each of the four reading modes.

In regard to the effect of regressions onreading rate, the individual data corroborate thegroup data. The data for eight of the subjectsdemonstrated that rate performance on a giventext was significantly and negatively correlatedwith the number of regressions effected . Thistends to confirm the idea prompted by the groupdata that slow readers tend to make more regres-sions than do fast readers.

The size of the regressions (regression mag-nitude) that an individual subject effects did notappear to be significantly correlated with readingrate in 7 out of 10 instances . This is the samefinding as reported for the group data. Hence,slow tactual readers may make more regressionsthan fast readers, but they do not necessarilymake larger regressions.

The reading process generally involves two

64


sequentially linked motor system operations.First, each line of text must be carefully scannedto enable perception and interpretation of thetextual material . Second, at the completion ofeach text line the reader must make a right-to-left return sweep to the beginning of the nextline. The duration of the line-changing operationhas herein been called "line-changing time ."Taenzer (18) reported a mean line-changing timeof 2.4 seconds for a single subject using theOptacon to read a long letterprint story (12 pagesof text). This 2.4 second line-changing time iswell within the range of the 1 .2 to 4.5 secondsreported here and is similar to the overall meanvalue of 3 .0 seconds for our group of 10 subjects.

Regarding reading rate and line-changing timeperformance, net reading rate for Optacon/letterprint reading has been found to be nega-tively correlated with line-changing time for thegroup data (see Fig . 6) . It is evident that whentotal reading time (including line-changing time)is used to calculate gross reading rate, one wouldexpect a correlation between rate and line-chang-ing time. However, in Figure 6 we excluded line-changing times to calculate net reading rate.There is then no a priori reason to expect acorrelation between these two variables . Line-changing time reflects a motor process with nocognitive component, whereas the time requiredfor left-to-right progress in reading reflects bothmotor and cognitive processes . The correlationbetween net reading rate and line-changing timeimplies that reading rate is, in part, determined(or constrained) by motor performance, withinthe context of the Optacon/letterprint readingtask.

Although there is a range of line-changingtimes across our subject population, the value ofline-changing time for a given subject appears tobe constant. When this observation is consideredalong with the evidence that individual Optaconreaders read at a constant rate over a wide rangeof text material (11), a more complete picture ofthe tactual Optacon reading process emerges.Specifically, a given Optacon reader invokes aconstant line-reading rate and a constant (albeitfaster) sweep rate to change lines . Consequently,the entire reading process appears to consist oftwo constant-rate phases: 1) a constant rate

cognitive plus motor component phase and 2) aconstant rate motor task phase.

The preceding discussion helps us to begin tounderstand the tactual reading process . How-ever, the observed difference in the level of read-ing performance as indicated by different ratesamong individual readers and different overallrates measured with different tactual systems,still remains to be fully explained . Moreover, thefactors that act to establish the individualizedlevel of motor system performance are notobvious.

SUMMARY

Reading rate is clearly related in part to thetime taken to change from one line of text to thenext and also to the number of regressions thatthe reader chooses to effect . Fast readers tend totake less time than slow readers to change fromone line to another and to make fewer regres-sions. These results prompt the idea that theremay be other factors contributing to individualtactual reading performance than limitations im-posed by a given tactual reading system and itsassociated tactual reading code . Finally, the simi-larity of tactual reading text-scanning patternsfor the several modes of tactual reading that wereinvestigated suggests that the observed differ-ence in rates is the only important distinctionbetween the several tactual reading modalities.

ACKNOWLEDGMENTS

We thank Harvey Lauer and Leonard Mowinski of theBlind Center at the Hines Veterans Administration Hospitalfor their considerable help in producing the embossed brailleand Inkbraille texts used in this study. We also thank Dr.Robert Stepp of the University of Illinois at Urbana-Champaign and David Holladay of (Raised Dot Computing)Lewisburg, PA, who graciously provided much of thesoftware used to produce Inkbraille.

REFERENCES

BEN ARY R: Touch, Typing in Ten Lessons . New York:Grosset Dunlap, 1963.BERTELSON P, MOUSTY P. : Modes operatoires clans lalecture de 1'ecriture braille . Le 7avail Hama? n 45 :13-23, 1982.BLISS JC, KATCHER MN, ROGERS CH, SHEPARD RP:Optical-to-tactile image conversion for the blind . IEEE

1.

2.

3 .

65


Trans Man-Mc " ,

Sys MMS-11(1) : 58-65, 1970.4. BRUSCIA KE, L VINS N S: Predictive factors in Opta-

con music reading . J y Impair Blind 7 : 309-312,1982.

5. BERKLEN K : Touch Rec r 7,'iriit ,if the Blind (translatedby FK Merry, originally published 1917) . New York:American Foundation for the Blind, 1932.

6. DAVIDSON PW, WILES-KETTENVIANN M, HABER RN,APPELLE S: Relationship between hand movements,reading competence, and passage difficulty in braillereading . Neuropsychologia 18 : 629-635, 1980.

7. EATMAN PF: An analytic study of braille reading.Austin : University of Texas, 1942 . (Ph .D. thesis .)

8. FOULKE E: Reading braille . In : Tactual P rception : ASourcebook, W. Schiff and E . Foulke (e. 1\rw York:Cambridge University Press, 1982, o .,.

9. HAMPSHIRE B : Worl,~ r With „e ,ii, Study ofBraille as a Medium of Comrmrni.rirna . Bern, Swit-zerland : UNESCO, 1981.

10. HISLOP DW, ZUBER BL, TRIMBLE J: Characteristics ofblind reading and manual scanning with an electronicreading aid . Proceedings of the Fifth ' ii i, I Confer-ence on Rehabilitation Engineering . Bethesda, MD:Rehabilitation Society of North America, 1982, p . 115.

11. HISLOP DW, ZUBER BL, TRIMBLE J: Characteristics ofreading rate and manual scanning patterns of blindOptacon reader. Hum Factors 25(4) : 379-389, 1983 .

12. HISLOP DW, ZUBER BL, TRIMBLE J: Inkbraille : A pilotstudy of a new reading medium for blind persons . J VisImpair Blind 78 : 326, 1984.

13. IRWIN RB: The war of the dots . In : As I Saw P . NewYork: American Foundation for the Blind . 1955.

14. KUSv.IIMA T : Visual Reading and Brad', Reading : AnErp< or -riled Investigation of the P! y>'iI gy and

l rogu of Visual and Tactual Pi rrl r . New York:American Foundation for the Blind, 1971.

15. MILLER GR, COLEMAN EB : A set of thirty-six nropassages calibrated for complexity . J Verb Lean

bBehav 6 : 851-854, 1967.

16. MILLER GR, COLEMAN EB : The measurement ofreading speed and the obligation to generalize to apopulation of reading materials. J Read Behav 4: 48-56, 1971.

17. SCHOOF LT: An analysis of Optacon usage . Res BollAm Found Blind 19 : 33-50, 1975.

18. TAENZER JC: An information processing model forvisual and tactile reading . Perception 1 : 147-160, 1972.

19. TERZIEFF I, STAGG V, ASHCROF” I' SC: Increasing read-ing rates with the Optacon: a pilot study. J Vis ImpairBlind . 76 : 17-22, 1982.

20. ZUBER BL, WETZEL PA: Eye movement determinantsof reading rate . In : Models of Oculornotor Behaviorand Control, BL Zuber (ed .) . Boca Raton, FL: Chemi-cal Rubber Co., 1981, pp . 193-208 .

Documents

Text-scanning patterns of blind readers using Optacon and braille