#AUTHOR: Challman, Beth E.
#TITLE: VARIABLES INFLUENCING THE IDENTIFICATION OF SINGLE BRAILLE CHARACTERS
#ORGANIZATION: University of Louisville
#CATEGORY: Reading, Braille, Recognition
#PUBLICATION: Thesis for MA Degree
#ABSTRACT: Recognition times were determined for 55 single braille characters. The data provided by this study were compared to that reported in an earlier similar study. There was little agreement between the findings of the two studies relative to the effects of certain stimulus characteristics on identification time. The lack of correspondence was attributed to the different ways in which identification was defined in the two studies. It was argued that the data reported in the earlier study would more appropriately be called a measure of stimulus registration, whereas the data gathered for this study represented identification time which included stimulus registration and processing time. Since there was a positive relationship between identification times and errors, it was suggested that practice with character identification to enhance reading performance might better be based not only on misidentified characters but on characters which are less readily identified.
CHAPTER I
Even though the braille code has enabled visually impaired individuals to gain information and pleasure by reading, braille reading has serious disadvantages. One disadvantage is the distressingly slow reading rate that characterizes the typical braille reader. A reasonable estimate of overall average reading rate is 90 words per minute, although reading rates reported by various investigators (Ethington, 1956; Foulke, 1962; Lowenfeld & Abel, 1967; Lowenfeld, Hatlen, & Abel, 1969; Meyers, Ethington, & Ashcroft, 1958; Nolan, Morris, Kederis, Fieg, & Smith, 1966) reveal a wide range. Factors which influence reading rates include the type of material to be read, the age and educational level of the reader, the amount of reading in which the subject ordinarily engages, the age at onset of blindness, and the age at which the subject learns to read braille, and so forth.
Josephson (1961) studied the leisure activities of blind adults and found that braille reading was infrequently mentioned by these adults as a pleasurable leisure activity. A related report on blind readers (Josephson, 1964) revealed that many blind readers engaged in braille reading only when there was no alternative access to the information they needed. It seems plausible that the reading rates of many blind individuals are so. painfully slow, these persons may prefer to read by listening to the recorded discs, tapes, and cassettes that are prepared for them. Others may resort to a passive acceptance of whatever information and pleasure is available to them by way of radio or television. We do not know for certain that individuals who read little or no braille are, in fact, slow readers. However, we do know reading as slowly as 90 words per minute for long periods of time is, at best, tedious. What accounts for the typically slow braille reading rate has yet to be determined.
From the earliest construction of a tactual code, attempts have been made to equate tactual reading with visual reading. Although the physical stimuli of the reading materials are encountered via different sensory modalities, the assumption is the reading process, visually or tactually, is basically the same. Consequently, braille text is patterned after printed text. The braille reader scans the line of braille from left to right with his finger as his sighted counterpart scans print visually, and the lines of braille are read from the top of the page to the bottom just as printed text is read.
The patterning of tactual reading after visual reading has influenced investigators to design experiments concerning braille reading using, in large part, materials and paradigms similar to those used in visual reading studies. For example, Harley and Rawls (1971) studied approaches to teaching braille reading that utilized the Scott, Foresman basal reader series (Robinson, Monroe, & Artley, 1965) which stresses whole word learning and Lippincott's Basic Reading series (McCracken & Walcott, 1963) which stresses a phonics approach. The problems encountered by these experimenters in preparing the reading materials, and by the teachers in implementing the various programs prescribed by the experimental design, illustrate difficulties that arise from direct transcription of printed texts into braille.
The Lippincott texts were found to contain words for which the braille reader lacks basic concepts, such words as "arab," "bog," "blimp," and "bale." Also, words such as "went" followed "wet," emphasizing the short e sound, and subjects found the "en" sign difficult to recognize. The "com" sign was missing in the series except for the word "come," and the "sh" sign occurred only in "she." Singular word forms were found on the vocabulary lists whereas the plural forms were printed in the text. Sometimes learning a new phonics rule required learning a new symbol.
The experimental design used by Harley and Rawls (1971) called for texts to be prepared in Grade I braille, Grade II braille, and in Initial Teaching Alphabet (ITA) symbols. In preparing either of the texts using the ITA alphabet, the 44 symbols required sometimes made spacing patterns uneven. There were fewer problems with the Scott, Foresman materials than with the Lippincott materials, although in both series, direct transcription from print to braille made the presentation of critical symbols random rather than sequential as they are for the sighted learner. The teachers found that in order to make the learning experience appropriate for blind readers, extensive supplemental materials were needed. This was especially true when the phonemic braille code conflicted with the system used in the printed text.
The subjective reports of the teachers indicated a higher degree of independence in reading and writing, fewer frustrations than are usually encountered, and above average word attack skills for those groups using the Lippincott series. The Scott, Foresman ITA group was also described as being more independent and showing more of an inclination to read voluntarily. In general, teachers reported the children made the transition from ITA code symbology to Grade II braille with ease, and were able to write as soon as they learned the symbols associated with the sounds. There was little reported enthusiasm on the part of the teachers for using Grade I braille, although it probably was more amenable to the phonics approach. Teachers reported slower progress on the part of the students who learned Grade I braille, especially those children they classified as immature and slow learners. However, a positive evaluation was that Grade I braille readers made better spellers, since writing and reading seems to be reciprocally reinforcing.
The post test results reported by Harley and Rawls (1971) revealed that neither the whole word approach nor the phonics approach showed overall superiority. However, when compared to sighted norms, those students using Grade II braille in either approach performed as well or better than their sighted peers, whereas the Grade I braille and ITA groups did not. However, the superior performance of the Grade II braille groups to the Grade I and ITA groups may have been an artifact of the experimental condition, since the post test paragraphs were presented in Grade II braille. Children who read Grade I braille or used the ITA alphabet were given a rather short period of time in which to make the transition to Grade II braille. Also, since the teachers all reported preparing extensive supplementary materials, it is not clear whether the performance of the various groups was a function of the approaches, the symbology, or the supplementary materials. It might even have been due to extra attention given to reading and related activities by both the students and the teachers.
Rex (1971) made an extensive study of various basal readers and also reported problems similar to those encountered by Harley and Rawls (1971). Rex (1971) reported that teachers seriously questioned the use of basal readers prepared for sighted children, especially at the preprimer level. The teachers she questioned all expressed a need for supplementary materials, and reported they usually prepared their own materials for the preprimer level. Accordingly, Rex (1971) analyzed several basal reading series in order to prepare supplementary materials which might be appropriate for use in conjunction with a majority of texts currently available. She reported that in the basal reading series she analyzed, changes continued to occur with subsequent revisions. She noted there was less control over vocabulary as more and more instruction in phonics was introduced, and that the student is presumed to be able to analyze words more efficiently, and is, therefore, put on his own at a faster pace. The increasing emphasis placed on phonics skills creates problems for the reader using the braille code since the development of phonics skills within the basal reading series prepared for sighted readers is often distorted by the use of braille contractions.
Influenced by the patterning of braille reading after that of the sighted community, and limited by the dearth of materials prepared particularly for blind students, teachers of braille have favored approaches to teaching reading that parallel the approaches used by teachers of language skills for sighted students. When Lowenfeld, Abel, and Hatlen (1969) questioned teachers in residential schools for the blind and in day school programs, they found that two-thirds of the teachers started braille reading instruction with the word or sentence method, and one-third began reading instruction with the alphabet. A large percentage of those teachers who did begin reading instruction using the letter method switched to the word method soon after the letters were learned. They used the whole word method, probably because it was in vogue for use with sighted students at the time. Cardinale (1973), who made a similar study four years later, reported few teachers taught braille as a subject. Fifty-eight percent of those teachers responding to the questionnaire reported they used the whole word approach, but many teachers reported they used the whole word approach only as it related to whole word signs. Cardinale's questionnaire revealed that a high percentage of teachers reported they used a variety of procedures, particularly for beginning reading instruction. The lower percentage of teachers reporting the use of the whole word method in the Cardinale (1973) study from the percent age reported in the Lowenfeld, Abel, and Hatlen (1969) study probably reflects the shift to the phonics approach being used in the sighted community. It seems plausible that the high percentage of teachers who reported using a variety of methods may very well reflect attempts to cope with the peculiar problems encountered when braille symbology conflicts with the sequencing of phonics skills in basal readers prepared for sighted children.
The acceptance of procedures modeled from procedures employed in the visual community has gone relatively unchallenged, although there are some researchers who suggest that the braille reader is engaging in a task quite dissimilar to that of the sighted reader. The visual reader, they contend, is able to perceive a word or group of words simultaneously and then proceed from segment to segment of the printed material with a speed that is influenced by experience, eye fixation, concentration, and so forth. On the other hand, the manner in which the braille reader encounters his display is by progressing sequentially from one character to the next in a single word, holding the characters in memory until he has perceived enough characters to form a word percept and then holding these word percepts in memory to form phrases and sentences. For the braille reader, then, the perceptual unit or "window" is the single braille character, not the word or group of words as it is for the sighted reader. Nolan and Kederis (1969), after a series of nine studies to determine the perceptual factors in character and word recognition, suggest
"Since it appears that the character is the perceptual unit in reading and that word recognition is the result of sequential integration of information reflected by the characters, it seems realistic to emphasize character recognition as well as word recognition in the early stages of reading instruction." (p. 50)
Whether all braille readers perceive a word character by character has been called into question by Kilpatrick (unpublished research findings, 1974). He presented braille words to subjects by arranging for the words to pass under the fingertip at various speeds. Braille characters were embossed on paper tape and when the tape passed beneath the fingertip at a speed which allowed subjects to read at a rate of 130 words per minute or faster, they reported that the words were perceived as whole patterns. When the tape speed was such that the reading rate fell below 130 words per minute, subjects reported that reading became difficult. Kilpatrick hypothesized that the slower tape speeds apparently effected a disintegration of word patterns, and subjects were forced to read a letter at a time. The controversy surrounding sequential character by character reading as opposed to whole word patterning has yet to be resolved.
On the assumption that the braille reader does process each character sequentially, a number of studies have been made which relate to the legibility of the braille code and to various factors which may influence legibility. An early investigation was done by the Uniform Type Committee (UTC) (1913). Their effort was to order the characters of the braille alphabet in terms of legibility. They constructed 20 lists of 160 characters per list. Among these 160 items were four presentations of each of the 25 alphabetic characters not under test. The remaining 60 items on each list were instances of the specific character of interest. The characters on each list were presented in random order. Each list was read by a subject and his reading time and errors were recorded. Differences in the time taken to read the lists were assumed to be due to differences in the legibility of the braille pattern under study. Using list reading time as a measure of legibility, the investigators were able to rank characters in terms of difficulty. The data revealed that the number of dots in a given character affected the recognition of that character; that is, those lists containing test characters with fewer dots were read faster.
An analysis of the errors made by the subjects in the UTC (1913) study showed that most of the errors involved confusing characters with similar configurations. Other errors were classified as missed dot errors, added dot errors, and orientation errors, such as reversals.
Nolan and Kederis (1969) also studied factors influencing the perception of single braille characters. Their method, however, determined the minimum exposure time required for identification of each of the 55 braille dot patterns that stand for a letter or small group of letters, or for various punctuation marks. They used a device called a tachistotactometer which permitted the observation of characters for controlled intervals of time.
The braille characters were embossed on special plasticized strips of paper mounted on a platform beneath the display surface of the instrument. The display surface contained holes corresponding to dot positions in a line of blank braille cells. In order to present the patterns under investigation, the platform beneath the display surface was elevated so that the characters embossed on the plastic sheet were raised through the holes of the display surface to the height of conventional braille dots. The subject was allowed to examine a single pattern for a prescribed amount of time after which the platform was lowered, thus removing the braille pattern on display. There were seven slips of plasticized paper, six of which contained eight braille characters randomly assigned to each slip. The seventh slip contained the remaining seven characters, making a total of 55 patterns presented.
Thirty-six elementary and high-school students served as subjects. There was, for each subject, a different random assignment of the braille characters embossed on the seven slips of plasticized paper, and a different random order for presenting the slips.
There were four 1-hour sessions during which time recognition thresholds were obtained using a modified method of limits procedure. Initially, all of the characters presented in a given session, usually 16, were exposed at an interval of time presumed to be too brief to permit identification. Subsequently, the interval was increased by .01 second until the subject had identified the character correctly in three out of four successive presentations. The minimum exposure time for criterion identification was defined as threshold for a given character. The subject was encouraged to make an identification each time a pattern was presented and instructed to move his finger to a position on the display surface that would permit the examination of the next character when it was presented. When a character had been identified to criterion, the presentation of the remaining characters on the sheet was changed from a sequential to a haphazard order to enable control for position effects among characters on a given sheet.
Each subject's identification threshold was defined as the time for the first of four exposures during which the character was identified correctly three times. By taking the mean of all subject's identification thresholds for each character, Nolan and Kederis were able to establish relative legibility values for the 55 characters under investigation. A rank order correlation of these values with the order of characters in the UTC study which were common to both investigations yielded a coefficient of .68 indicating moderate overall agreement with the earlier legibility ranking.
With respect to the number of dots in a given pattern, Nolan and Kederis found, as in the UTC study, that character identification time was positively related to the number of dots in a character. The spatial arrangement of dots was also a subject of analysis in the Nolan and Kederis study. They found that among characters with the same number of dots, those characters with more widely dispersed dots were more quickly identified, as were those with dots falling in the top two rows of a cell.
An analysis of incorrect responses revealed that 86% of the errors were due to missed dots. There were more missed dot errors for patterns with dots in the bottom row of the cell than in characters with dots in the top row. More dots were missed in the right column than in the left column of the cell. The investigators suggested that the apparent tendency of the braille reader to attend more closely to the upper left-hand portion of the pattern may be, at least in part, a consequence of the direction of hand motion when braille is read. They also suggested, on the basis of an earlier study, certain features of the braille patterns as they are used to represent written English, occur with higher probabilities.
Kederis, Siems, and Haynes (1965) made a frequency count of the occurrence of the 63 characters in the braille code that appeared in various materials prepared for blind readers. They also counted the occurrences of the dots appearing in the six dot locations within those cells. They found that dots occurring on the left side of the cell occur 7% more frequently than dots on the right side of the cell, and dots in the upper part of the cell were 8% more prevalent than those in the bottom part. A comparison of the missed dot errors in the Nolan and Kederis study with frequency of occurrence counts obtained by Kederis et al. (1965) yielded an inverse relationship between the two sets of data. Accordingly, Nolan and Kederis concluded that the perception of a braille character is based, in part, on the expectancy of what dots occur most frequently.
In an additional study, Nolan and Kederis (1969) compared identification times for braille words with the sum of the identification times for the individual characters of which the words were composed. In this study, subjects were classified as slow and fast readers on the basis of timed reading tests. The apparatus and procedure were the same as in the previously cited study by these same authors. The major relevant findings with respect to single character legibility were (a) for faster readers, character identification thresholds were significantly lower than were those for slow readers; (b) the order of legibility of braille characters was, in general, the same for both groups; and (c) the fast and slow readers did not appear to differ with respect to the types of errors made.
In addition to the data reported for single character legibility, Nolan and Kederis (1969) found that the time needed to identify a word frequently exceeded the sum of the identification times of the letters in that word. This is in contrast to the behavior of visual readers who can recognize words of varying lengths at exposure times shorter than or equivalent to the sums of the exposure times of individual letters in those words. Foulke and Wirth (1973), for example, found that for sighted subjects, the minimum exposure time needed to identify a word was frequently no greater than the minimum exposure time for a single letter in that word.
Nolan and Kederis (1969) also compared the recognition time of various words with the amount of time it took blind readers to encounter sequentially all of the characters in those words. They found, in general, that word recognition time was longer than the time it took a blind subject to pass his finger over the word. The results of this study led Nolan and Kederis to conclude the blind reader must be processing his display character by character. When the blind reader identifies a word, it takes him longer to do so than it takes him to identify all of the individual characters of which the word is composed because he must identify sequentially each character, hold these identifications in memory until he has processed enough to form a word precept. The visual reader, on the other hand, can identify whole words in less time than it takes to identify all the individual alphabetic characters because his scanning procedure is entirely different. Foulke and Wirth (1973) conducted an experiment in which printed text was displayed to visual readers a letter at a time and found that under these conditions, the visual reading speed is comparable to that of a braille reader. In an earlier study, Troxell (1967) also found that the average visual reading speed was 19.5 words per minute when text was presented a letter at a time, and the average tactual reading speed was 18 words per minute when the display was read character by character.
Nolan and Kederis (1969) assumed, then, that the perceptual unit is the individual braille cell and that training in character recognition to increase speed and accuracy might have an effect on braille reading efficiency. Accordingly, they conducted a study to test this assumption. Three pretest measures were taken from 12 control subjects selected from Grades 3 through 6, and from 12 experimental subjects selected from the same grades. These pretest measures included a silent reading rate and a comprehension test score from a Gates Basic Reading Test, an oral reading rate and error score from a 400-word, unfamiliar short story, and an oral reading rate and error score from a character identification test composed of the 55 single braille characters previously studied.
Experimental subjects were given training for ½ hour for 18 consecutive days to increase rate and accuracy of character identification. The training consisted of drill on characters misidentified in the pretest and of work on various discrimination tasks. After training, post-test measures were taken using alternate forms of the pretest. Subjects' time for and errors in character identification, and oral reading errors all decreased significantly while their oral reading speeds increased significantly. However, there was not a significant increase in silent reading rate nor in scores on the comprehension test covering the material read silently. A related experiment with elementary students done by Henderson (1967) yielded similar results. Nolan and Kederis (1969) and Henderson (1967) indicated variables unassociated with the reading process per se, namely motivational variables associated with extrinsic rewards, may have influenced the results. Umsted (1970) trained high-school students in character identification and reported an increase in reading rate from 97 words per minute to 126 words per minute.
Although the research on character legibility apparently provided the rationale for the research reported on the effects of character identification practice on reading efficiency, in all three of the research efforts mentioned, difficulty of character identification was defined as those characters which were misidentified during pretesting. It does seem plausible to assume that misidentified characters are those which will impede reading of connected prose, and to, therefore, spend time practicing on those characters. However, it is tempting to conjecture that practice on characters having low legibility, defined by longer minimum exposure times needed for identification, might very well influence reading rate and comprehension, and that even practice on those characters defined as having high legibility might also effect reading ability.
If the perceptual unit for the braille reader is the single braille character, as Nolan and Kederis suggest, and if braille reading speed depends on the speed and accuracy with which a braille reader identifies these characters, then training which increases speed and accuracy of identification should produce more efficient braille readers. The data suggest that training does reduce errors and increase speed of identification and that the facilitating effects from training in character identification transferred to the oral reading of connected discourse.
In the Nolan and Kederis (1969) study in which identification thresholds were determined, the measure accepted as threshold was the minimum exposure time permitting absolute identification. A modified psychophysical method of limits procedure was used and it seems more appropriate to consider the threshold values thus obtained as a measure of stimulus registration. However, the task was not a true psychophysical task in the sense that a subject is reporting, for instance, he either hears a tone of a given amplitude and frequency or he does not hear it. The subject not only reported he felt a dot pattern beneath his finger-tip, he reported what that particular dot pattern represented. In other words, he operated on his sensory experience, upon the sensory registration he no longer experienced, and reported he had made the appropriate discriminations and had constructed a meaningful perception. The measure permitting identification did not include what might be called processing time. However, the training given in character identification was based on measures which included stimulus registration and processing time. Such a measure might be called identification threshold, and the measure reported by Nolan and Kederis might be termed, more appropriately, stimulus registration threshold.
Certain stimulus characteristics were found to affect stimulus registration thresholds, and the question being raised now is whether those same stimulus characteristics would also affect character identification thresholds, where the duration of the stimulus is under the control of the subject, as it is in the typical reading experience. In other words, is processing also a function of legibility? If so, then training based on identification time data, such as those data used by Nolan and Kederis, Henderson, and Umsted, would be indicated. If, on the other hand, processing time is stable, depending upon stimulus characteristics, then training to lower stimulus registration thresholds may be indicated.
Haber and Nathanson (1969) using visual stimuli, found that processing time, defined as onset to onset of sequentially presented stimuli, predicted recognition better than the duration of the stimuli. The implication of this finding is that once exposure time is sufficient for registration, further increases in on-time are irrelevant if no masking conditions occur during processing time. If one can plausibly extend this paradigm to tactual perception, perhaps the blind reader cannot read faster because subsequent letters constitute masking interfering with the serial processing. If the processing of a given character is not yet complete when the next letter is encountered, then the registration of the next character would interfere with the processing of the incompletely processed letter. If this were the case, then practice to reduce the time for registration of the character might increase the time available for the processing that follows stimulus registration, thereby decreasing the likelihood that the registration of the next character would interfere. If, however, practice cannot reduce stimulus registration times due to characteristics inherent in the code, and if processing time is also a function of these same stimulus characteristics, it may well be that teaching strategies that encourage the student to respond to word configurations rather than to single character patterns would be indicated. If such were the case, practice to increase vocabulary and decrease word recognition times should affect reading performance positively.
There is some evidence to support the contention that experience with words tends to shift the reader from a letter by letter strategy to a whole word strategy. Doggett and Richards (1975) studied the effect of word length on recognition thresholds using sighted subjects. They found that word length was influential in lengthening recognition thresholds for words that appear less frequently in English prose. They hypothesized that if a word occurs infrequently in language, the subject may not have it in his lexical memory, and therefore, must construct the word letter by letter. However, if a subject is experienced in language, as he is presumed to be with words occurring with high frequency, then word length not only has no effect on recognition time, but may even act as a cue, decreasing recognition time. In an earlier study, Newbigging and Hay (1962) also reported a strong word length effect for low frequency words. In their study, they used college students as subjects, and found the word length effect more pronounced for freshmen and sophomores than for upperclassmen. They hypothesized that their upper level college students were less susceptible to word length effects on recognition, be cause their recognition vocabularies may have been larger than those of the lowerclassmen. One would, of course, need a follow-up study to determine if word length effects decreased when the lower level students presumably acquired more experience with infrequently occurring words.
The hypothesis is tempting, however, since it seems to support the findings of those who work with students characterized as learning disabled. It is frequently reported that these individuals fall farther and farther behind their peers as they advance through the school grades, and since this is the case, they cannot gain experience with the increasing vocabulary load, and therefore may be forced to read using the letter by letter strategy. It may be assumed that any word with which the student is unfamiliar, no matter how frequently it occurs in his language, is a low frequency item for him, and therefore not in his lexical store.
It is, however, beyond the scope of this paper, to enter into the controversy over whole word recognition and serial character processing. The data, to date, appear to support the serial processing of individual braille characters.
In the present study, measures of what have been referred to as identification time were determined for 55 of the braille code characters used by Nolan and Kederis in their study. The purpose of the study was to investigate whether the same variables which affected the identification time thresholds reported by Nolan and Kederis (1969) would also affect identification time data when the duration of the stimulus was under the control of the subject rather than determined by the experimenter. In addition, due to the limitations imposed by the tachistotactometer used by Nolan and Kederis, they found it unfeasible to completely randomize presentations. The tachistotactometer designed for this study allowed for the correction of this methodological problem.
CHAPTER II
Method
Subjects. Ten experienced braille readers served as subjects. All were either living in the city or were temporarily residing at the Kentucky Rehabilitation Center for the Blind in Louisville, Kentucky. Table I provides for each subject the age, educational level, occupation, age at onset of blindness, approximate amount of hours per week spent in reading braille, and the type of material usually read.
Apparatus and materials. A tachistotactometer designed and constructed at the Perceptual Alternatives Laboratory, a graduate research institute of the University of Louisville, was used to present the stimuli. The electronic components were housed in a wooden box .46 x .43 x .22 meters. A small metal display surface on the top of the machine was drilled with six holes corresponding to the six dot positions of a single braille cell. In order to present a braille pattern, pins below the display surface were elevated through the holes to the height of conventional braille dots. Each of the six pins was activated by its own solenoid and since the mass of a pin was small and it did not have to travel far when it rose above the surface, the presentation was immediate. Similarly, when the pins were lowered beneath the display surface, the removal of the stimulus to prevent further inspection was also immediate. Table II gives the rise times and retraction times for each of the pins. The times reported are not identical, but differences are negligible. Only one dot pattern at a time was available for inspection, therefore the problems of position effects or order effects were precluded.
A small box housing six toggle switches and one master switch was connected to the tachistotactometer. Each of the toggle switches operated one of the pin solenoids, and the master switch connected on the common return of all six solenoids was used to present the dot pattern that was selected by closing the appropriate switches. The master switch also turned on a Hunter timer, Model 1522, which could be read in milliseconds. Each character which was presented remained elevated and the timer continued to run until the subject announced his identification into a microphone. The microphone signal generated by the subject operated a switch which stopped the timer and turned off the power supply to the solenoids, allowing the pins that specified dot patterns to recede below the display surface.
The tachistotactometer was mounted at desk height on a rack in such a way that it could be adjusted for the comfort of each participant. Since there were no adjacent patterns to cue the subject as to where he should place his finger, a small finger guide was mounted on the display surface to ensure proper placement of the finger to be stimulated.
For ease in selecting the pattern to be presented and for recording the data, a chart outlining the orders of presentation for each subject was prepared. The 55 characters were assigned randomly to each of 10 trials for each subject.
Procedure. Each subject was briefly interviewed to obtain the information presented in Table I. The subject was encouraged to examine the equipment and was given informal instructions about the experimental task. Each subject was asked to read silently a short prose passage.
The experimenter indicated when the subject should begin reading and the subject notified the experimenter when he had finished reading the entire passage. A stopwatch was used to time the silent reading so that the subject's reading rate could be determined. The experimenter asked ques tions about the passage so that she could determine whether the subject had paid attention to what had been read.
The subject was then seated comfortably before the tachistotactometer with the directional microphone placed about 6 inches from his mouth. A few pretrials were given to familiarize the subject with the task and to allow him to learn to announce his identifications loud enough to activate the voice key. The experimenter then indicated to the subject that the experimental session would begin. When the subject's preferred hand was in place, the experimenter consulted the appropriate chart, set up the indicated pattern by throwing the appropriate toggle switches, and threw the master switch to present the stimulus. Response latency in milliseconds and errors were recorded. Interstimulus intervals were not controlled. Stimuli were presented as rapidly as possible.The experimenter varied the setting up period so that the subject could not predict from the amount of time it took the experimenter to set up a pattern whether a one- or two-dot character was probable, or a five- or six-dot pattern was likely to occur. After the first 55 character trial was completed, the experimenter proceeded to the next trial without comment, to preclude the subject's predicting the occurrence of a character from the new set of 55.
The number and duration of experimental sessions varied depending on the time a given subject had available. If sessions became long, subjects were allowed a 5- to 10-minute break. Each session was scheduled at the subject's convenience and any session was terminated when a subject became fatigued, bored, or had no more time to give. The data were collected at the Perceptual Alternatives Laboratory and at the Kentucky Rehabilitation Center for the Blind.
CHAPTER III
Results
Median identification times, in milliseconds (msec), were determined from the 10 identifications each subject reported for each character. A mean of these medians was computed for each of the 55 characters. The range of identification times was .87 - 1.64, with a mean of 1.23. The mean identification time for alphabetic characters was 1.13 and for contractions 1.86 Table III presents the mean identification times for the 55 single braille characters.
A rank order correlation of mean identification times obtained in this study with the measures of stimulus registration obtained by Nolan and Kederis (1969) yielded a .30 coefficient, indicating relatively small agreement with the earlier legibility ranking. Since the variability between subjects appeared to be considerable, the Kendall W, a statistic of ranking concordance (Kendall, 1948) was computed. The coefficient of concordance over the 55 characters was not statistically significant (W = .04), indicating that subjects apparently do not rank the 55 characters similarly. However, coefficients of concordance computed for only the alphabetic characters (W = .34) and for only contraction characters (W = .20) were both significantly different from zero, indicating there is some agreement among subjects as to the ranking of characters within either category.
A one-way analysis of variance (Winer, 1962) was done to test whether the identification times for alphabetic characters differed significantly from those to contraction patterns. The difference was significant (p < .01), with the alphabetic characters identified more readily than the contraction patterns. This finding is in agreement with the apparent subject concordance relative to rankings within categories which was not evident over all characters. Table IV presents the results of this analysis, F .01 (1, 9) = 10.56.
Two more simple analyses of variance were done to determine the effects of the number of dots in character configurations on identification time, and the effect of stimulus ambiguity on response latency. Ambiguity is defined as identical dot configurations appearing in differing cell locations. The results of both analyses indicated that neither variable, number of dots nor ambiguity, produced any significant effects on identification times.
It seemed probable that individuals who identify characters more readily are faster readers than those individuals with longer identification times. However, a correlation between the reading rates of the 10 subjects and their 55 character mean identification times indicated a negligible inverse relationship (r = -.14). Since the mean of alphabetic characters differed significantly from that of contraction characters, two more correlations were done to determine whether the identification times for either category were related to reading speed. Again the correlations (r = .13 and r = -.15) revealed a negligible relationship between how fast an individual identifies characters and how fast he reads. Table V gives the reading rates of each subject, and their identification times averaged over 55 characters, and over contractions and alphabet symbols.
Since subjects kept the stimulus under the fingertip until they made their identifications, and since the instructions did not encourage subjects to respond until they could identify the stimulus pattern, it was not anticipated there would be many, if any, errors. However, 379 errors were recorded. Erroneous identifications of alphabetic symbols occurred 111 times, and contraction errors occurred almost three times more frequently, 276 times.
In an attempt to determine whether an error prone subject is also a slow reader, a correlation was done between reading speeds and the numbers of errors made by each of the 10 subjects. The correlation yielded a coefficient (r = -.49), indicating a moderate inverse relationship between the two measures. However, the coefficient was not found to differ significantly from zero.
A correlation between the mean identification times for each of the 55 characters and the number of errors made on those characters yielded a product moment coefficient of .42. The statistic was significantly different from zero (p< .005), indicating a strong relationship between the amount of time it takes subjects to identify characters and the errors made of those characters.
In order to determine the factors which might contribute to misidentification, errors were grouped into five categories: omission errors added dot errors, position errors, orientation errors, and other errors. An omission error is made when the subject identifies a character that is identical to the stimulus except it contains one less dot than the stimulus configuration. An added dot error is made by the subject when his response includes all the dots in the stimulus and one more dot. A position error is made when the subject perceives the correct configuration of dots but he incorrectly perceives the position in the cell. Orientation errors are defined as instances in which the identified character represents a rotation of the stimulus pattern around either the X, Y, or Z axis. Those errors which fall in none of the above categories are classified as other errors. Table VI gives the percentage of errors in each category.
Omission errors were made more frequently than added dot errors, and dots occurring in the bottom of the cell were missed more frequently than those occurring in the top portion of the cell. Left cell dot omissions accounted for 56% of omission errors, and dots occurring in the right portion of the cell were missed 45% of the time. Dots occurring in the lower portion of the cell were missed more frequently than those occurring in the upper portion of the cell. Table VII gives the percentage of omission errors in each of the six cell locations.
A correlation between the identification times for alphabetic characters and the frequency of occurrence counts made by Kederis et al. (1965) revealed an inverse relationship between the two measures (r = -.36), which was significantly different from zero (p < .05). A similar correlation between contraction identification times and the frequency with which they occur in braille reading materials also revealed an inverse relationship (r = -.23). Although it might be expected that less frequently occurring items would take longer to identify, the relationship indicated by the coefficient was not statistically significant.
A correlation between errors made on alphabetic characters with the frequency of occurrence of those characters and a similar correlation between contraction errors and frequency of occurrence yielded product moment coefficients of .005 and -.303, respectively, neither of which were significantly different from zero. Likewise, the correlation be tween the frequency with which dots appear in each of the six cell locations and the number of times those dots were omitted, yielded a statistic (r = .223) not significantly different from zero.
CHAPTER IV
Discussion
If there is any general statement that can be made about the perceptual ability of individuals to identify tactual patterns, it would probably include some consideration of the discriminable features of the pattern. In the present study, it was anticipated that the same features of braille patterns which affected the data gathered by Nolan and Kederis (1969) would also influence these data. In the present study, the amount of time available for examining the pattern under test was controlled by the subject, whereas in the earlier study (Nolan & Kederis, 1969), stimulus duration was determined by the experimenter. However, in either case, subjects were required to make absolute identifications of the 55 single braille characters, and it was anticipated there would be some agreement between the two measures.
One pattern feature which can be presumed to influence the time needed to identify a pattern is complexity, here defined as the number of dots in the test pattern. In this study, there was no significant effect on identification time due to complexity, in contrast with the earlier study (Nolan & Kederis, 1969) in which the number of dots was found to be positively related to the stimulus duration needed in order to make an identification.
Another pattern feature which was found to effect stimulus registration time in the Nolan and Kederis study was pattern ambiguity. Those investigators found that longer exposure times were needed to permit the identification of characters with dots occupying the lower two rows of the cell than were needed for identical characters in which dots occupied the two top rows of the cell. The results of the present study fail to indicate any effects of pattern location on identification. It appears that the subjects participating in this study were as likely to quickly identify complex, ambiguous characters with dots in the upper two rows of a cell as they were to respond to simple, unambiguous patterns with dots in the lower two rows of the cell.
Although identification times were not affected by ambiguity, accuracy in identifying ambiguous stimuli was influenced. Fifty-two percent of all incorrect identifications were location errors. Subjects may respond quickly to ambiguous stimuli, but they make errors, especially when there are no contextual cues to inform them.
The second largest percentage of errors were those classified as omission errors. Nolan and Kederis (1969) reported 86% of all errors made by their subjects were missed dot errors. In the data gathered for this study, omission errors accounted for 28% of the misidentifications. Dots were more frequently omitted from the bottom row of the cell than from the top row, which is in agreement with the finding reported by Nolan and Kederis. However, contrary to the data reported by the earlier investigators, in this study more dots were omitted from the left column of the cell than from the right column. These data do not support the hypothesis that the reader acquires certain inclinations to attend to the upper left portion of the pattern, since he typically scans the line of braille from left to right as he reads. Neither is there support for the hypothesis that the braille reader develops expectancies of what dots are likely to occur with greater frequency.
An unexpected finding was the absence of a relationship between speed of character identification and reading rate. Since there was a significant difference between the response time means of alphabetic characters and contraction patterns in favor of the alphabetic symbols, it seemed possible that contractions are more problematic for readers. However, the correlations of reading rate with identification time for either letter symbols or contraction symbols were negligible. It is interesting to note, though, the fastest reader was not one who identified characters quickly; while the slowest reader had, in general, the shortest identification times. The slower reader was an individual whose vision had deteriorated sufficiently to require that she learn to read braille, and she had been reading braille for only 2½ months. Since she had recently mastered the braille code and was developing skill as a braille reader, it may have been she was attending more closely to the individual symbols as she read. The faster reader, on the other hand, was a proofreader of braille materials and she had considerable braille reading experience. If the role of experience tends to shift the attention of the reader to larger and larger segments of the material, then the faster reader may have been reading in just that manner. This can only be conjecture, since there has not been sufficient experimental investigation to provide information concerning what a skilled braille reader does when he reads. It is, however, highly probable that the faster reader had acquired considerably more experience predicting the occurrence of items and combinations of items in braille text. As was pointed out in Chapter I, braille contractions do not always correspond to the phonemic conventions of printed material, and the slower reader had considerably more practice with print reading than with braille symbology.
There was a modest negative correlation between reading speed and errors. Although the relationship is modest, it does indicate an effect in the expected direction. One does not need to argue the unlikelihood that an error prone reader is either a good reader or a fast reader.
It was expected that the braille characters under test would be ranked similarly to the legibility order reported by Nolan and Kederis (1969). However, the two sets of measures were virtually uncorrelated. This is not entirely surprising. Even though both studies required pattern identifications, their measures were of two different time intervals. It may be the rankings would have corresponded more had there been a measure of what has been called stimulus registration in this study, to compare with that measure in the Nolan and Kederis study. Additionally, the apparatus used for this study was such that it allowed for measurements in milliseconds, whereas the tachistotactometer used for the earlier study measured in hundredths of seconds, therefore there were many tied ranks in those data.
The correlation of most interest is that between response times and identification errors. If legibility, or character difficulty, is defined in terms of relative identification times or in terms of relative numbers of errors on those characters, there is agreement. There also seems to be agreement among the subjects concerning the ranking of contraction items and the ranking of alphabetic items as well. The Kendall W statistic for either of those categories was significantly different from zero. It is not surprising there are differences between the two categories, since investigators and teachers of braille consistently report contractions as the troublemakers for braille readers.
The question, raised earlier in this paper, concerned whether practice on braille characters to enhance typical reading performance should be based on legibility data or on identification time data. Since this investigation did not determine what has been called stimulus registration thresholds, there can be no consideration of the theoretical implications of those data with respect to practice effects. Studies which have involved practice have used error data to indicate which code items are to receive attention. Since there is a correlation between identification times and errors, it may do well to practice for speed as well as accuracy. In the typical reading task, symbols are available for examination until an identification is made. When a reader must spend time puzzling over the identification, or when he makes errors, his reading skill is deficient.
In remedial programs for learning disabled sighted subjects, an effort has been made to drill students in initial sounds, vowel sounds, blends, endings, and so forth, to both accuracy and speed criteria. In addition, vocabulary lists, connected discourse, and spelling are all practiced until the responses are fast as well as accurate. Not only does practice improve speed and accuracy, but there are reported gains in comprehension of silently read material.
Whether the braille reader can approximate, tactually, the reading skill exhibited by a good sighted reader, is an empirical question, the answer to which awaits investigations to resolve the serial character integration/whole word controversy, and studies to determine what a good braille reader does when he reads.
Summary
The purpose of the study reported in this paper was to investigate the effects of various pattern characteristics on the recognition times for single braille cells. A prior investigation by Nolan and Kederis (1969) ordered 55 braille characters in terms of difficulty. Difficulty was defined as the minimum exposure time needed for absolute correct identification of the braille symbol. In the present study, characters were ordered according to the mean of 10 subjects' median response times to each of these same 55 characters. There was little correlation between the rank orders determined in the two studies. A Kendall W statistic of concordance revealed that there was subject variability when all 55 characters were ranked, but within the categories of alphabetic characters and contraction characters, the concordance for either category was statistically significant. A simple analysis of variance revealed that the identification time means of contractions and alphabetic characters differed significantly, with the shorter identification times for alphabetic characters. No effects of dot number on identification times were found. This is contrary to the findings in the earlier study. Neither were there effects of stimulus ambiguity on identification measures. However, misidentifying identical stimulus configurations that could occur in different cell locations, accounted for 52% of all errors.
Omission errors, or missed dot errors, accounted for 28% of all errors in this study, as opposed to 82% reported in the comparison study. In agreement with the earlier data, more dots were omitted from the bottom row of the cell than from the top row. Contrary to the earlier findings, in this study more dots were missed on the left side of the character than in the right column. There is some doubt that braille readers attend more to the upper left portion of the cell because braille is typically read from left to right.
There was no relationship between reading rate and identification times, but there was a very modest negative correlation between reading rate and identification errors. A statistically significant relationship was found between identification time and identification errors. Since other studies have reported modest improvement in reading speed and accuracy after practice was given on misidentified characters, and since there was a positive correlation between response times and errors, it was suggested that practice to both speed and accuracy criteria be considered.
LIST OF REFERENCES
Cardinale, J. F. Methods and procedures of braille reading. American Foundation for the Blind Research Bulletin, 1973, 26, 171-183.
Doggett, D., & Richards, L. G. A reexamination of the effect of word length on recognition thresholds. American Journal of Psychology, 1975, 88, 583-594.
Ethington, D. The readability of braille as a function of three spacing variables. Unpublished master's thesis, University of Kentucky, 1956.
Foulke, E. The role of experience in the formation of concepts. International Journal for the Education of the Blind, 1962, 12, 1-6.
Foulke, E., & Wirth, E. M. The role of identification in the reading of braille and print. In E. Foulke (Ed.), The development of an expanded reading code for the blind: Part II. Louisville: Perceptual Alternatives Laboratory, University of Louisville, 1973.
Haber, R. N., & Nathanson, L. S. Processing of sequentially presented letters. Perception and Psychophysics, 1969, 5(6), 359-361.
Harley, R. K., & Rawls, R. F. Comparison of several approaches for teaching braille reading to blind children. American Foundation for the Blind Research Bulletin, 1971, 23, 63-85.
Henderson, F. M. The effect of character recognition training on braille reading. Unpublished specialist in education thesis, George Peabody College for Teachers, Nashville, 1967.
Josephson, E. A report on blind readers. New Outlook for the Blind, 1964, 58, 97-101.
Josephson, E. Some insights concerning reading problems as reflected in AFB leisure activity study. Report of Proceedings of Conference on Research Needs in Braille. New York: American Foundation for the Blind, Inc., 1961.
Kederis, C. J., Siems, J. R., & Haynes, R. L. A frequency count of the symbology of English braille grade 2, American usage. International Journal for the Education of the Blind, 1965, 15, 38-46.
Kendall, M. G. Rank correlation methods. London: Charles Griffin & Co., Ltd., 1948.
Kilpatrick, J. F. Increasing the braille reading rate by presenting connected prose written in braille at controlled rates. In E. Foulke (Ed.), Annual Report to the Dean of the Graduate School Concerning the Perceptual Alternatives Laboratory. Louisville: Perceptual Alternatives Laboratory, University of Louisville, 1974.
Lowenfeld, B., & Abel, C. L. Methods of teaching braille reading: Final report. San Francisco, Calif.: San Francisco State College, 1967.
Lowenfeld, B., Abel, C. L., & Hatlen, P. H. Blind children learn to read. Springfield: Charles C. Thomas, 1969.
McCracken, C., & Walcott, C. C. Basic reading. Philadelphia: J. B. Lippincott Co., 1963.
Meyers, E., Ethington, D., & Ashcroft, S. C. Readability of braille as a function of three spacing variables. Journal of Applied Psychology, 1958, 42, 163-165.
Newbigging, P. L., & Hay, J. The practice effect in recognition thresholds as a function of word frequency and length. Canadian Journal of Psychology, 1962, 16, 177-184.
Nolan, C. Y., & Kederis, C. J. Perceptual factors in braille word recog nition. New York: American Foundation for the Blind, Research Series #20, 1969.
Nolan, C. Y., Morris, J. E., Kederis, C. J., Fieg, K. E., & Smith, M. C. Reading and listening in learning by the blind: Progress report. Louisville: American Printing House for the Blind, 1966.
Rex, E. J. A study of basal readers and experimental supplementary materials for teaching primary reading in braille. Education of the Visually Handicapped, 1971, 3(1), 1-7.
Robinson, H. M., Monroe, M., & Artley, A. S. The new basic reading curriculum foundation series. Chicago: Scott, Foresman & Co., 1965.
Troxell, D. E. Experiments in tactile and visual reading. IEEE Trans actions on human factors in electronics (HFE-8), 1967, 261-263.
Umsted, R. C. Improvement of braille reading through training. Unpublished doctoral dissertation, George Peabody College for Teachers, Nashville, 1970.
Uniform Type Committee of the American Association of Workers for the Blind. Fourth Biennial Report. (Reprinted in Outlook for the Blind, 1913, 7, 1-48.)
Winer, B. J. Statistical principles in experimental design. New York: McGraw-Hill Book Co., 1962.