- Email: [email protected]
Visual information processing in relation to age and to reading ability
JOURNAL
OF EXPERIMENTAL
Visual Information
CHILD
PSYCHOLOGY
27, 143-152 (1979)
Processing in Relation Reading Ability
to Age and to
JOHN L. ARNETT The University of Manitoba AND
VINCENT
DI LOLLO
The l7niversit.v of Alberta Temporal aspects of early visual information processing were studied developmentally in good and in poor reading male subjects ranging in age from 7 to 13 years. Forced-choice temporal integration and backward masking tasks, respectively, were utilized to assess duration of visual persistence and of relative rate of visual information processing. The results did not reveal differences in either visual persistence or processing rate in relation to reading ability at any age level studied. However, processing rate was found to increase markedly with chronological age in both the good and the poor readers while visual persistence did not vary significantly. The findings were discussed in relation to earlier work and in relation to current theoretical formulations of visual information processing.
Initial perceptual stages in the reading process involve abstraction and subsequent interpretation of information derived from a number of discrete eye fixations on the text. Each successive fixation feeds new information into the visual system which must be integrated with earlier information in the development of meaning. In view of the sequential nature of the incoming stimulation, temporal aspects of early visual information processing are of obvious importance in the study of reading and have potential significance in the etiology of reading disorders. Exploration of the relationship between reading disorders and the initial stages of visual information processing may be directed at differences in This research was supported by the National Research Council of Canada, Grant A0269, to the second author. Requests for reprints should be sent to Jobn L. Arnett, Department of Psychiatry, Faculty of Medicine, The University of Manitoba, 770 Bannatyne Avenue, Winnipeg R3E OW3, Canada. 143 Copyright @ 1919 by Academic Press, Inc. All rights of reproduction in any form reserved.
144
ARNETT
AND
DI LOLL0
duration of visual persistence and at the relative rates of information processing in groups of poor and of normal readers. Stanley and Hall (1973) attempted to estimate the duration of visual persistence in poor and in normal readers by measuring the interstimulus interval (ISI) at which two sequentially presented portions of a composite display appeared to separate. They also examined relative processing rates in the same subjects by utilizing a backward-masking paradigm where the perception of a temporally leading alphabetic character was impaired by a trailing mask composed of an aggregate of dots. Poor readers were found to have longer visual persistence and a slower rate of processing, as indexed by performance on temporal integration and on backward-masking tasks, respectively. However, Fisher and Frankfurter (1977) provided contrasting results which indicated that poor readers in fact processed visual information more rqidy than normal-reading controls. In addition they suggested that their results possibly indicated longer visual persistence in normal readers. The basis for the differences between these two sets of findings remains unclear. It is conceivable that the ascending method of limits employed by Stanley and Hall (1973) may have resulted in biased estimates of visual persistence and of processing rates in that poor readers may have adopted a more conservative response criterion. Given the poor readers’ past history of failure in testing situations, it appears reasonable that they may have waited longer to be sure before indicating when the two-part composite stimuli appeared to separate in Stanley and Hall’s temporal integration task as well as before reporting the target letter in the backward masking paradigm. Furthermore, the above research failed to explore visual persistence and processing rate developmentally despite evidence that these aspects of visual information processing vary with age (Blake, 1974; Gummerman & Gray, 1972: Liss & Haith, 1970; Miller, 1972; Welsandt, Zupnick, & Meyer, 1973). The present research was designed to examine the duration of visual persistence and the relative processing rate in poor and in normal readers. To avoid potential confounding by response-criterion differences, forced-choice methodol’ogy with nonverbal stimuli was used. The experimental paradigm was similar to that of Eriksen and Collins (1968) who presented two sequential displays which, although meaningless individually, formed a meaningful trigram when temporally integrated. Eriksen and Collins (1968) regarded the estimates obtained with this paradigm as representing the duration of a short-term visual store (Neisser, 1967). METHOD
~u&x~.Y. Forty-eight male students, 12 at each of four age levels (7, 9, 1 I, and 13 years) from Metropolitan Winnipeg participated as subjects. Six subjects at each age level were poor readers (experimental subjects)
VISUAL
PROCESSES
AND
READING
145
ABILITY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.
0
0
.
.
0
0
.
0
0
0
0
.
.
.
0
0
0
0
0
0
0
0
0
0
0
l
0
0
0
0
.
0
0 FIG.
0 I.
0
a Display
matrices
and fixation
point.
and six were normal-reading controls. All had normal hearing and vision, were of at least average intelligence, attended school regularly, and were native English speaking. All experimental subjects performed at least one year below grade level in reading on the basis of teacher evaluations and of scores on the Stanford Achievement Test. All control subjects performed above grade level. Stimuli u& uppurutus. The stimulus display consisted of two horizontally adjacent 5 x 5 square dot matrices each measuring 1.O cm square and separated by 0.5 cm (Fig. I). The display was viewed binocularly through a Tektronix model 016-0154-00 viewing hood. At the viewing distance of approximately 75 cm the display subtended visual angles of l”53’ horizontally and 0’46’ vertically. The 25 dots forming each of the two matrices were plotted on a Tektronix 602 display oscilloscope equipped with ultrafast PI5 phosphor. One dot was omitted from one of the matrices on each trial. The incomplete matrix (left or right) was chosen randomly on each trial. The precise location of the missing dot within the incomplete matrix varied randomly from trial to trial in one experimental condition (temporal integration) but was fixed at the centre of the matrix in the other condition (backward masking). A dim fixation dot, located 0.25 cm from the inner edges and 0.5 cm from the top and bottom boundaries of the matrices, was employed to aid and to standardize subject orientation to the screen. The fixation dot disappeared at the onset of the stimulus display and reappeared following a response. The luminance of all displays was maintained at a constant level throughout the experiment. Periodic calibration of the oscilloscope ensured that a standard square test patch plotted on the screen yielded a reading of 0.35 lux as measured by a Tektronix Jl6 digital photometer. To aid focusing and convergence the oscilloscope display surface was dimly illuminated. All displays were generated by a PDP-8/L computer which also performed all timing and scoring functions. Proceduw. The subject was seated in a quiet and dimly illuminated room and was instructed to focus on the fixation dot and to initiate a trial when ready by depressing a foot switch. FoIlowing the stimulus display,
146
ARNETT
AND
DI LOLL0
he was required to depress a switch which he held in his left hand if he thought that the dot was missing from the left matrix or to depress a switch which he held in his right hand if it appeared to him that the dot was missing from the right matrix. Immediately following the subject’s response the fixation dot reappeared and the same sequence of events proceeded for the next trial. The subject was instructed to work carefully and as quickly as possible without sacrificing accuracy and to make a best guess when unsure of the location of the missing dot. Prior to the beginning of the experiment all subjects were familiarized with the display conditions and with the task requirements. During this period all subjects demonstrated virtually errorless performance in identifying the incomplete matrix. Although ail stimuli were displayed at a visual angle that placed them entirely within the fovea, it might have been possible to perform the experimental task by focusing attention on just one rather than on both matrices. The subjects were carefully instructed to attend to both matrices and, indeed, debriefing at the end of the experiment confirmed that this strategy had been employed by every subject. Data were collected on two experimental tasks, temporal integration and backward masking, as explained below. Tenzporul infegruGon. Successful performance at identifying the incomplete matrix depends on the simultaneous perceptual availability of all 49 dots. In the present task the dots were plotted sequentially over time. Hence some form of sensory persistence was necessary to bridge the temporal gaps between the dots. It is known (Hogben & Di Lollo, 1974; Di Lollo, 1977) that when all the dots forming the matrix elements are presented sequentially within a brief temporal interval they appear clearly and simultaneously. At longer intervals, however, the dots presented early in the plotting sequence appear to be missing and are easily confused with the truly missing dot. Thus, as the total time that elapses between plotting the first and the last dots, known as the p/offing infervul, is increased, identification of the incomplete matrix becomes progressively more difficult. The present temporal integration task dynamically varied the duration of the plotting interval until a desired level of accuracy in identifying the incomplete matrix was achieved. Every dot in each matrix (except the missing dot) was plotted once only for 1.5 microseconds. Dots were plotted in pairs, one from each matrix; each member of a pair was chosen randomly and independently with respect to the 25 possible locations in each matrix. Dot pairs were evenly spaced within the plotting interval. A computer-controlled psychophysical tracking procedure developed by Taylor and Creelman (1967) known as PEST (Parameter Estimation by Sequential Testing) was employed to adjust dynamically the duration of the plotting interval to a level which yielded approximately 80% correct
VISUAL
PROCESSES
AND
READING
ABILITY
147
responses. It must be noted that successive dot-pairs remained evenly spaced irrespective of the duration of the plotting interval. A run began with the computer randomly selecting an initial plotting interval ranging between 3 and 127 msec. A series of trials was conducted at that plotting interval and the computer maintained a record of the subject’s performance. In order to determine whether the subject’s performance was above or below an accuracy criterion of 80% correct, a Wald (1947) sequential likelihood-ratio test, integrated with the PEST program. was performed on the data. Accuracy above 80% correct was considered to reflect a task which was too easy (too brief a plotting interval), while accuracy below 80% correct indicated that the task was too difficult (too long a plotting interval). If the task was too easy (or too difficult) the PEST program automatically increased (or decreased) the duration of the plotting interval. The magnitude of each adjustment made to the duration of the plotting interval decreased as the 80% accuracy level was approached. The run terminated when an adjustment in the plotting interval was required which was smaller than a minimum step value which had been set at 8 msec. The duration of the plotting interval corresponding to the 80% performance level was recorded at the end of each run. Backward rnaskirzg. This part of the study employed a stimulus display that was fundamentally the same as the one employed in the temporal integration task with two exceptions. First, the unplotted dot was always randomly missing from the centre of either the left or the right matrix. And, second, all the dots forming the two matrices were presented effectively simultaneously for a duration of 3 msec. The matrix display was followed, after a blank interstimulus interval (ISI), by a backward-masking stimulus of approximately the same luminance and duration as the matrix display. The masking stimulus consisted of 50 dots positioned randomly within the spatial confines of the two matrices. In the present study the IS1 between the matrix display and the masking stimulus was dynamically adjusted by the PEST procedure described earlier. A run began with the computer randomly selecting a starting IS1 ranging between zero and 127 msec, and it ended when an IS1 adjustment smaller than 12 msec was required. In both temporal integration and backward masking tasks a series of eight runs, each lasting approximately 3 min, was conducted per subject with a brief rest period following each run. Half the subjects at each age and reading level performed the temporal integration task on one day and the backward masking task on the following day while the remaining subjects performed the tasks in the reverse order. RESULTS
The median of the eight scores obtained from each subject in the temporal integration and in the backward masking tasks, respectively,
148
ARNETT
AND
DI
LOLL0
was computed separately for the two tasks. To examine the possibility of systematic differences in the range of eight scores obtained by each subject, each of the eight scores was expressed as a percentage of the mean score of that particular subject in the temporal integration and in the backward masking tasks, separately. No systematic differences attributable either to age or to reading ability were obtained in the range of scores. Every subject thus had two scores: one for the temporal integration task and one for the backward masking task. The temporal integration score represented the median duration of the plotting interval at which the subject achieved approximately SOY5 correct responses; the backward masking score denoted the median IS1 at which the 8% accuracy level was met in the backward masking task. These scores formed the basis for all analyses and summaries performed on the data. Separate analyses of variance were performed on the scores for the two tasks. Each was an entirely between-subjects analysis comprising two levels of reading ability (good and poor readers), two levels of task performance order (temporal integration first or second), and four levels of chronological age (7, 9, 11, and 13 years). Figure 2 shows mean scores on the backward masking task for good and for poor readers, separately, at each of the four age levels, averaged over task performance order. It is clear that the IS1 required to escape the disruptive effects of the mask decreased markedly as chronological age increased [F(3,32) = 22.3, p < .OOl]. However, neither reading abihty nor any other main effect or interaction significantly affected the results (F < 11. Figure 3 shows mean scores on the temporal integration task for good and for poor readers, separately, at each of the four age levels, averaged over task performance order. None of the main effects or interactions were significant (F < 1). DISCUSSION
Normal and deficient readers did not differ significantly from one another in performance on either the temporal integration or backward masking tasks at any of the four age levels studied. Performance on temporal integration tasks has typically been regarded as an index of the duration of visual persistence (Eriksen 8z Collins, 1%8; Hogben & Di Lollo, 1974; Stanley & Hall, 19731. In a sequentially presented multielement composite display visual persistence is necessary in order to bridge the temporal gap between the element. Namely, success at such tasks depends on the simultaneous perceptual availability of all parts of the display even though presented sequentially. On this basis, the results illustrated in Fig. 3 strongly suggest that poor readers did not differ from normal readers in the duration of visual persistence. Performance on backward-masking tasks is widely regarded as an index
VISUAL
PROCESSES AND READING
lG0
-
EXPERIMENTAL
o---o
CONTROL
I 7
I 9
149
ABILITY
I 11
I 13
AGE (YEARS)
FIG. 2. evel.
Mean IS1 (msec) between test stimulus and mask as a function of age and reading
70 t
M
EXPERIMENTAL
0 1
9
,
,
11
13
AGE (YEARS1
FIG. 3. Mean plotting interval (msec) as a function of age and reading level. The plotting interval represented the total time that elapsed between plotting the first and last dots (i.e., the total duration of the display). Ah dots were evenly spaced within a given plotting interval, regardless of its duration.
150
ARNETT
AND
DI LOLL0
of the rate of visual information processing (Blake, 1974; Gummerman & Gray, 1972; Liss & Haith, 1970; Spitz & Thor, 1968). Masking may occur in at least two ways. The contours of test and masking stimuli may integrate into a composite image from which the test stimulus cannot be extracted. Impairment occurs if such integration takes place before the viewer has had sufficient time to identify the test stimulus. Alternatively, the onset of the mask may imerrupt ongoing test-stimulus processing either by abrogating processing mechanisms required commonly by the two displays or by causing a rapid shift in attention. As has been compellingly argued by Scheerer (1973) backward masking is likely to occur through perceptual integration at relatively brief stimulus-onset asynchronies (SOA) and by interruption of processing at longer SOAs. Whether backward masking occurs through integration or through interruption, impairment in performance is the result of insufficient time to process the test stimulus before the onset of the mask. The data in Fig. 2 clearly show that younger children required a longer mask-free interval to reach the same criterion-level of performance as older children and hence displayed the slower rate of information processing. This interpretation is buttressed by the entirely consonant pattern of results reported by Welsandt et ~1. (1973). Employing a backward masking task, these investigators found that performance of young children did not differ from the adult level for ISIS up to about 50 msec but that marked superiority of adult performance developed at longer ISIS. However, no differences in relative processing rates were found between good and poor readers at any of the age levels studied (Fig. 2). This result points to the improbability that the basis for reading disabilities might be found at such early stages of information processing as might be affected by backward masking. This conclusion agrees entirely with the position of Vellutino, Steger, Kaman, and De Setto (1975). It must be realized, however. that the nature of the matrix stimuli and the response requirements in the present study permitted an assessment of visual information processing only at a relatively superficial level of analysis. That is, the experimental tasks employed did not necessarily involve any verbal or semantic encoding which clearly is involved in reading proper and is generally thought to occur at higher processing levels (Mackworth, 1972). It appears quite likely that at a deeper level of processing differences between good and poor readers would emerge. This line of reasoning follows from the Craik and Lockhart (1972) “levels of processing” conceptual framework which suggests that preliminary levels of processing are more concerned with the analysis of physical or sensory features of the stimuli while deeper levels involve to a greater degree matching the input against stored abstractions from past learning. In this model, greater “depth” implies a greater degree of semantic or cognitive analysis. Hence
VISUAL
PROCESSES
AND READING
ABILITY
151
it is possible that, under stimulus conditions and task demands more similar to those actually involved in reading, differences between good and poor readers would emerge in processing rate. Our results and conclusions stand in disagreement with the pattern of results reported by Stanley and Hall (1973) as well as with the opposing pattern of results reported by Fisher and Frankfurter (1977). Stanley and Hall (1973) presented evidence indicating longer visual persistence and slower rates of visual information processing in poor readers relative to normal reading control children. Interestingly, and in support of the present contention that the findings were affected by response-criterion differences, Stanley (1976) subsequently employed a forced-choice approach in a digit identification task under backward masking conditions with poor and normal readers. Under these circumstances, his previous finding of more rapid processing by the normal-reading control subjects was not replicated. Turning to the results of Fisher and Frankfurter (1977), there is no apparent theoretical rationale for the findings of more rapid processing by reading-retarded children. Nor did the authors provide an explanation for these unusual findings. Furthermore, the evidence implying longer visual persistence in normal readers appears rather weak and, by the authors’ own admission, only suggestive. One final point needs to be made. At first glance it may be suggested that the present tasks might have been too gross to be sensitive to differences in sensory persistence or in processing rates that might exist between good and poor readers. However, this hypothesis does not appear tenable when it is realized that while no differences between normal and deficient readers emerged at any of the age levels studied, the results clearly show that the rate of information processing increased markedly with increasing chronological age. It should also be noted that the direct relationship between chronological age and processing rate found in the present study is entirely consistent with other studies relating these two variables in normal-reading children (Gummerman & Gray, 1972: Liss & Haith, 1970; Miller, 1972; Welsandt et al., 1973). Furthermore, an inspection of the backward masking scores revealed that all the 7-year-old subjects processed information more slowly than any of the 13-year-old subjects, regardless of reading ability. Thus, the backward masking task was highly sensitive to different developmental levels in both experimental and control subjects and would be expected to be equally sensitive to differences in rate of processing ascribable to factors other than chronological age, notably reading deficiencies. REFERENCES Blake, J. Developmental Journal
of Experimental
change in visual information Child
Psychology,
1974.
processing under backward masking. 17, 133-146.
152
ARNETT
AND
DI LOLL0
Craik, F. I. M., & Lockhart, R. S. Levels of processing: A framework for memory research. Journal
of Verbal
Learning
and
Verbal
Behavior,
1972,
11, 671-684.
Di Lollo, V. Temporal characteristics of iconic memory. Nature, 1977, 267, 241-243. Eriksen, C. W., & Collins, J. F. Sensory traces versus the psychological moment in the temporal organization of form. Journal of Experimenta/ Psycho/ogy, I%& 77,376-382. Fisher, D. F., & Frankfurter, A. Normal and disabled readers can locate and identify letters: Where’s the perceptual deficit? Journal of Reading Behavior, 1977, 9, 31-43. Gummerman, K., & Gray, C. Age, iconic storage and visual information processing. Journa/ of Experimental
Child
Psychology,
1972,
13, 165-170.
Hogben, J. H., & Di Lollo, V. Perceptual integration and perceptual segregation of brief visual stimuli. Vision Research, 1974, 14, 1056-1069. Liss, P. H., & Haith, M. M. The speed of visual processing in children and adults: Effects of backward and forward masking. Percepfion and Psychophysics, 1970, 8, 396-398. Mackworth, J. F. Some models of the reading process: Learners and skilled readers. Reading Research Quarterly, 1972, 7, 701-733. Miller, L. K. Visual masking and developmental differences in information processing. Child Development, Neisser, U. Cognitive
Spitz, H. H., &Thor, and
1972, 43, 704-709. psycho/ogy. New York:
Appleton-Century-Crofts, 1967. D. H. Visual backward masking in retardates and normals. Perception
Psychophysics,
1968, 4, 245-246.
Scheerer, E. Integration,
interruption and processing rate in visual backward masking. 1973, 36, 71-93. Stanley, G. The processing of digits by children with specific disability (Dyslexia). Brirish Psychologische
Journal
Forschung,
of Educational
Psychology,
1976, 46, 81-84.
Stanley, Cl., & Hall, R. Visual information processing in dyslexics. Chi/d Deve/opment, 1973, 44, 841-844. Taylor, M. M., & Creelman, C. D. PEST: Efficiency estimates on probability functions. The Journal of rhe Acoustical Sociery of America, l%7, 41, 782-787. Vellutino, F. R., Steger, J. A., Kaman, M., & De Setto, L. Visual form perception in deficient and normal readers as a function of age and orthographic-linguistic familiarity. Correx, 1975, 11, 22-30. Wald, A. Sequenria/ ana/ysis. New York: John Wiley & Sons, 1947. Welsandt, R. F., Zupnick, J. J., & Meyer, P. A. Age effects in backward visual masking (Crawford paradigm). Journal of Experimenral Child Psychology, 1973, 15, 454-464. RECEIVED: January 3, 1978; REVISED: March 21, 1978.
OF EXPERIMENTAL
Visual Information
CHILD
PSYCHOLOGY
27, 143-152 (1979)
Processing in Relation Reading Ability
to Age and to
JOHN L. ARNETT The University of Manitoba AND
VINCENT
DI LOLLO
The l7niversit.v of Alberta Temporal aspects of early visual information processing were studied developmentally in good and in poor reading male subjects ranging in age from 7 to 13 years. Forced-choice temporal integration and backward masking tasks, respectively, were utilized to assess duration of visual persistence and of relative rate of visual information processing. The results did not reveal differences in either visual persistence or processing rate in relation to reading ability at any age level studied. However, processing rate was found to increase markedly with chronological age in both the good and the poor readers while visual persistence did not vary significantly. The findings were discussed in relation to earlier work and in relation to current theoretical formulations of visual information processing.
Initial perceptual stages in the reading process involve abstraction and subsequent interpretation of information derived from a number of discrete eye fixations on the text. Each successive fixation feeds new information into the visual system which must be integrated with earlier information in the development of meaning. In view of the sequential nature of the incoming stimulation, temporal aspects of early visual information processing are of obvious importance in the study of reading and have potential significance in the etiology of reading disorders. Exploration of the relationship between reading disorders and the initial stages of visual information processing may be directed at differences in This research was supported by the National Research Council of Canada, Grant A0269, to the second author. Requests for reprints should be sent to Jobn L. Arnett, Department of Psychiatry, Faculty of Medicine, The University of Manitoba, 770 Bannatyne Avenue, Winnipeg R3E OW3, Canada. 143 Copyright @ 1919 by Academic Press, Inc. All rights of reproduction in any form reserved.
144
ARNETT
AND
DI LOLL0
duration of visual persistence and at the relative rates of information processing in groups of poor and of normal readers. Stanley and Hall (1973) attempted to estimate the duration of visual persistence in poor and in normal readers by measuring the interstimulus interval (ISI) at which two sequentially presented portions of a composite display appeared to separate. They also examined relative processing rates in the same subjects by utilizing a backward-masking paradigm where the perception of a temporally leading alphabetic character was impaired by a trailing mask composed of an aggregate of dots. Poor readers were found to have longer visual persistence and a slower rate of processing, as indexed by performance on temporal integration and on backward-masking tasks, respectively. However, Fisher and Frankfurter (1977) provided contrasting results which indicated that poor readers in fact processed visual information more rqidy than normal-reading controls. In addition they suggested that their results possibly indicated longer visual persistence in normal readers. The basis for the differences between these two sets of findings remains unclear. It is conceivable that the ascending method of limits employed by Stanley and Hall (1973) may have resulted in biased estimates of visual persistence and of processing rates in that poor readers may have adopted a more conservative response criterion. Given the poor readers’ past history of failure in testing situations, it appears reasonable that they may have waited longer to be sure before indicating when the two-part composite stimuli appeared to separate in Stanley and Hall’s temporal integration task as well as before reporting the target letter in the backward masking paradigm. Furthermore, the above research failed to explore visual persistence and processing rate developmentally despite evidence that these aspects of visual information processing vary with age (Blake, 1974; Gummerman & Gray, 1972: Liss & Haith, 1970; Miller, 1972; Welsandt, Zupnick, & Meyer, 1973). The present research was designed to examine the duration of visual persistence and the relative processing rate in poor and in normal readers. To avoid potential confounding by response-criterion differences, forced-choice methodol’ogy with nonverbal stimuli was used. The experimental paradigm was similar to that of Eriksen and Collins (1968) who presented two sequential displays which, although meaningless individually, formed a meaningful trigram when temporally integrated. Eriksen and Collins (1968) regarded the estimates obtained with this paradigm as representing the duration of a short-term visual store (Neisser, 1967). METHOD
~u&x~.Y. Forty-eight male students, 12 at each of four age levels (7, 9, 1 I, and 13 years) from Metropolitan Winnipeg participated as subjects. Six subjects at each age level were poor readers (experimental subjects)
VISUAL
PROCESSES
AND
READING
145
ABILITY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.
0
0
.
.
0
0
.
0
0
0
0
.
.
.
0
0
0
0
0
0
0
0
0
0
0
l
0
0
0
0
.
0
0 FIG.
0 I.
0
a Display
matrices
and fixation
point.
and six were normal-reading controls. All had normal hearing and vision, were of at least average intelligence, attended school regularly, and were native English speaking. All experimental subjects performed at least one year below grade level in reading on the basis of teacher evaluations and of scores on the Stanford Achievement Test. All control subjects performed above grade level. Stimuli u& uppurutus. The stimulus display consisted of two horizontally adjacent 5 x 5 square dot matrices each measuring 1.O cm square and separated by 0.5 cm (Fig. I). The display was viewed binocularly through a Tektronix model 016-0154-00 viewing hood. At the viewing distance of approximately 75 cm the display subtended visual angles of l”53’ horizontally and 0’46’ vertically. The 25 dots forming each of the two matrices were plotted on a Tektronix 602 display oscilloscope equipped with ultrafast PI5 phosphor. One dot was omitted from one of the matrices on each trial. The incomplete matrix (left or right) was chosen randomly on each trial. The precise location of the missing dot within the incomplete matrix varied randomly from trial to trial in one experimental condition (temporal integration) but was fixed at the centre of the matrix in the other condition (backward masking). A dim fixation dot, located 0.25 cm from the inner edges and 0.5 cm from the top and bottom boundaries of the matrices, was employed to aid and to standardize subject orientation to the screen. The fixation dot disappeared at the onset of the stimulus display and reappeared following a response. The luminance of all displays was maintained at a constant level throughout the experiment. Periodic calibration of the oscilloscope ensured that a standard square test patch plotted on the screen yielded a reading of 0.35 lux as measured by a Tektronix Jl6 digital photometer. To aid focusing and convergence the oscilloscope display surface was dimly illuminated. All displays were generated by a PDP-8/L computer which also performed all timing and scoring functions. Proceduw. The subject was seated in a quiet and dimly illuminated room and was instructed to focus on the fixation dot and to initiate a trial when ready by depressing a foot switch. FoIlowing the stimulus display,
146
ARNETT
AND
DI LOLL0
he was required to depress a switch which he held in his left hand if he thought that the dot was missing from the left matrix or to depress a switch which he held in his right hand if it appeared to him that the dot was missing from the right matrix. Immediately following the subject’s response the fixation dot reappeared and the same sequence of events proceeded for the next trial. The subject was instructed to work carefully and as quickly as possible without sacrificing accuracy and to make a best guess when unsure of the location of the missing dot. Prior to the beginning of the experiment all subjects were familiarized with the display conditions and with the task requirements. During this period all subjects demonstrated virtually errorless performance in identifying the incomplete matrix. Although ail stimuli were displayed at a visual angle that placed them entirely within the fovea, it might have been possible to perform the experimental task by focusing attention on just one rather than on both matrices. The subjects were carefully instructed to attend to both matrices and, indeed, debriefing at the end of the experiment confirmed that this strategy had been employed by every subject. Data were collected on two experimental tasks, temporal integration and backward masking, as explained below. Tenzporul infegruGon. Successful performance at identifying the incomplete matrix depends on the simultaneous perceptual availability of all 49 dots. In the present task the dots were plotted sequentially over time. Hence some form of sensory persistence was necessary to bridge the temporal gaps between the dots. It is known (Hogben & Di Lollo, 1974; Di Lollo, 1977) that when all the dots forming the matrix elements are presented sequentially within a brief temporal interval they appear clearly and simultaneously. At longer intervals, however, the dots presented early in the plotting sequence appear to be missing and are easily confused with the truly missing dot. Thus, as the total time that elapses between plotting the first and the last dots, known as the p/offing infervul, is increased, identification of the incomplete matrix becomes progressively more difficult. The present temporal integration task dynamically varied the duration of the plotting interval until a desired level of accuracy in identifying the incomplete matrix was achieved. Every dot in each matrix (except the missing dot) was plotted once only for 1.5 microseconds. Dots were plotted in pairs, one from each matrix; each member of a pair was chosen randomly and independently with respect to the 25 possible locations in each matrix. Dot pairs were evenly spaced within the plotting interval. A computer-controlled psychophysical tracking procedure developed by Taylor and Creelman (1967) known as PEST (Parameter Estimation by Sequential Testing) was employed to adjust dynamically the duration of the plotting interval to a level which yielded approximately 80% correct
VISUAL
PROCESSES
AND
READING
ABILITY
147
responses. It must be noted that successive dot-pairs remained evenly spaced irrespective of the duration of the plotting interval. A run began with the computer randomly selecting an initial plotting interval ranging between 3 and 127 msec. A series of trials was conducted at that plotting interval and the computer maintained a record of the subject’s performance. In order to determine whether the subject’s performance was above or below an accuracy criterion of 80% correct, a Wald (1947) sequential likelihood-ratio test, integrated with the PEST program. was performed on the data. Accuracy above 80% correct was considered to reflect a task which was too easy (too brief a plotting interval), while accuracy below 80% correct indicated that the task was too difficult (too long a plotting interval). If the task was too easy (or too difficult) the PEST program automatically increased (or decreased) the duration of the plotting interval. The magnitude of each adjustment made to the duration of the plotting interval decreased as the 80% accuracy level was approached. The run terminated when an adjustment in the plotting interval was required which was smaller than a minimum step value which had been set at 8 msec. The duration of the plotting interval corresponding to the 80% performance level was recorded at the end of each run. Backward rnaskirzg. This part of the study employed a stimulus display that was fundamentally the same as the one employed in the temporal integration task with two exceptions. First, the unplotted dot was always randomly missing from the centre of either the left or the right matrix. And, second, all the dots forming the two matrices were presented effectively simultaneously for a duration of 3 msec. The matrix display was followed, after a blank interstimulus interval (ISI), by a backward-masking stimulus of approximately the same luminance and duration as the matrix display. The masking stimulus consisted of 50 dots positioned randomly within the spatial confines of the two matrices. In the present study the IS1 between the matrix display and the masking stimulus was dynamically adjusted by the PEST procedure described earlier. A run began with the computer randomly selecting a starting IS1 ranging between zero and 127 msec, and it ended when an IS1 adjustment smaller than 12 msec was required. In both temporal integration and backward masking tasks a series of eight runs, each lasting approximately 3 min, was conducted per subject with a brief rest period following each run. Half the subjects at each age and reading level performed the temporal integration task on one day and the backward masking task on the following day while the remaining subjects performed the tasks in the reverse order. RESULTS
The median of the eight scores obtained from each subject in the temporal integration and in the backward masking tasks, respectively,
148
ARNETT
AND
DI
LOLL0
was computed separately for the two tasks. To examine the possibility of systematic differences in the range of eight scores obtained by each subject, each of the eight scores was expressed as a percentage of the mean score of that particular subject in the temporal integration and in the backward masking tasks, separately. No systematic differences attributable either to age or to reading ability were obtained in the range of scores. Every subject thus had two scores: one for the temporal integration task and one for the backward masking task. The temporal integration score represented the median duration of the plotting interval at which the subject achieved approximately SOY5 correct responses; the backward masking score denoted the median IS1 at which the 8% accuracy level was met in the backward masking task. These scores formed the basis for all analyses and summaries performed on the data. Separate analyses of variance were performed on the scores for the two tasks. Each was an entirely between-subjects analysis comprising two levels of reading ability (good and poor readers), two levels of task performance order (temporal integration first or second), and four levels of chronological age (7, 9, 11, and 13 years). Figure 2 shows mean scores on the backward masking task for good and for poor readers, separately, at each of the four age levels, averaged over task performance order. It is clear that the IS1 required to escape the disruptive effects of the mask decreased markedly as chronological age increased [F(3,32) = 22.3, p < .OOl]. However, neither reading abihty nor any other main effect or interaction significantly affected the results (F < 11. Figure 3 shows mean scores on the temporal integration task for good and for poor readers, separately, at each of the four age levels, averaged over task performance order. None of the main effects or interactions were significant (F < 1). DISCUSSION
Normal and deficient readers did not differ significantly from one another in performance on either the temporal integration or backward masking tasks at any of the four age levels studied. Performance on temporal integration tasks has typically been regarded as an index of the duration of visual persistence (Eriksen 8z Collins, 1%8; Hogben & Di Lollo, 1974; Stanley & Hall, 19731. In a sequentially presented multielement composite display visual persistence is necessary in order to bridge the temporal gap between the element. Namely, success at such tasks depends on the simultaneous perceptual availability of all parts of the display even though presented sequentially. On this basis, the results illustrated in Fig. 3 strongly suggest that poor readers did not differ from normal readers in the duration of visual persistence. Performance on backward-masking tasks is widely regarded as an index
VISUAL
PROCESSES AND READING
lG0
-
EXPERIMENTAL
o---o
CONTROL
I 7
I 9
149
ABILITY
I 11
I 13
AGE (YEARS)
FIG. 2. evel.
Mean IS1 (msec) between test stimulus and mask as a function of age and reading
70 t
M
EXPERIMENTAL
0 1
9
,
,
11
13
AGE (YEARS1
FIG. 3. Mean plotting interval (msec) as a function of age and reading level. The plotting interval represented the total time that elapsed between plotting the first and last dots (i.e., the total duration of the display). Ah dots were evenly spaced within a given plotting interval, regardless of its duration.
150
ARNETT
AND
DI LOLL0
of the rate of visual information processing (Blake, 1974; Gummerman & Gray, 1972; Liss & Haith, 1970; Spitz & Thor, 1968). Masking may occur in at least two ways. The contours of test and masking stimuli may integrate into a composite image from which the test stimulus cannot be extracted. Impairment occurs if such integration takes place before the viewer has had sufficient time to identify the test stimulus. Alternatively, the onset of the mask may imerrupt ongoing test-stimulus processing either by abrogating processing mechanisms required commonly by the two displays or by causing a rapid shift in attention. As has been compellingly argued by Scheerer (1973) backward masking is likely to occur through perceptual integration at relatively brief stimulus-onset asynchronies (SOA) and by interruption of processing at longer SOAs. Whether backward masking occurs through integration or through interruption, impairment in performance is the result of insufficient time to process the test stimulus before the onset of the mask. The data in Fig. 2 clearly show that younger children required a longer mask-free interval to reach the same criterion-level of performance as older children and hence displayed the slower rate of information processing. This interpretation is buttressed by the entirely consonant pattern of results reported by Welsandt et ~1. (1973). Employing a backward masking task, these investigators found that performance of young children did not differ from the adult level for ISIS up to about 50 msec but that marked superiority of adult performance developed at longer ISIS. However, no differences in relative processing rates were found between good and poor readers at any of the age levels studied (Fig. 2). This result points to the improbability that the basis for reading disabilities might be found at such early stages of information processing as might be affected by backward masking. This conclusion agrees entirely with the position of Vellutino, Steger, Kaman, and De Setto (1975). It must be realized, however. that the nature of the matrix stimuli and the response requirements in the present study permitted an assessment of visual information processing only at a relatively superficial level of analysis. That is, the experimental tasks employed did not necessarily involve any verbal or semantic encoding which clearly is involved in reading proper and is generally thought to occur at higher processing levels (Mackworth, 1972). It appears quite likely that at a deeper level of processing differences between good and poor readers would emerge. This line of reasoning follows from the Craik and Lockhart (1972) “levels of processing” conceptual framework which suggests that preliminary levels of processing are more concerned with the analysis of physical or sensory features of the stimuli while deeper levels involve to a greater degree matching the input against stored abstractions from past learning. In this model, greater “depth” implies a greater degree of semantic or cognitive analysis. Hence
VISUAL
PROCESSES
AND READING
ABILITY
151
it is possible that, under stimulus conditions and task demands more similar to those actually involved in reading, differences between good and poor readers would emerge in processing rate. Our results and conclusions stand in disagreement with the pattern of results reported by Stanley and Hall (1973) as well as with the opposing pattern of results reported by Fisher and Frankfurter (1977). Stanley and Hall (1973) presented evidence indicating longer visual persistence and slower rates of visual information processing in poor readers relative to normal reading control children. Interestingly, and in support of the present contention that the findings were affected by response-criterion differences, Stanley (1976) subsequently employed a forced-choice approach in a digit identification task under backward masking conditions with poor and normal readers. Under these circumstances, his previous finding of more rapid processing by the normal-reading control subjects was not replicated. Turning to the results of Fisher and Frankfurter (1977), there is no apparent theoretical rationale for the findings of more rapid processing by reading-retarded children. Nor did the authors provide an explanation for these unusual findings. Furthermore, the evidence implying longer visual persistence in normal readers appears rather weak and, by the authors’ own admission, only suggestive. One final point needs to be made. At first glance it may be suggested that the present tasks might have been too gross to be sensitive to differences in sensory persistence or in processing rates that might exist between good and poor readers. However, this hypothesis does not appear tenable when it is realized that while no differences between normal and deficient readers emerged at any of the age levels studied, the results clearly show that the rate of information processing increased markedly with increasing chronological age. It should also be noted that the direct relationship between chronological age and processing rate found in the present study is entirely consistent with other studies relating these two variables in normal-reading children (Gummerman & Gray, 1972: Liss & Haith, 1970; Miller, 1972; Welsandt et al., 1973). Furthermore, an inspection of the backward masking scores revealed that all the 7-year-old subjects processed information more slowly than any of the 13-year-old subjects, regardless of reading ability. Thus, the backward masking task was highly sensitive to different developmental levels in both experimental and control subjects and would be expected to be equally sensitive to differences in rate of processing ascribable to factors other than chronological age, notably reading deficiencies. REFERENCES Blake, J. Developmental Journal
of Experimental
change in visual information Child
Psychology,
1974.
processing under backward masking. 17, 133-146.
152
ARNETT
AND
DI LOLL0
Craik, F. I. M., & Lockhart, R. S. Levels of processing: A framework for memory research. Journal
of Verbal
Learning
and
Verbal
Behavior,
1972,
11, 671-684.
Di Lollo, V. Temporal characteristics of iconic memory. Nature, 1977, 267, 241-243. Eriksen, C. W., & Collins, J. F. Sensory traces versus the psychological moment in the temporal organization of form. Journal of Experimenta/ Psycho/ogy, I%& 77,376-382. Fisher, D. F., & Frankfurter, A. Normal and disabled readers can locate and identify letters: Where’s the perceptual deficit? Journal of Reading Behavior, 1977, 9, 31-43. Gummerman, K., & Gray, C. Age, iconic storage and visual information processing. Journa/ of Experimental
Child
Psychology,
1972,
13, 165-170.
Hogben, J. H., & Di Lollo, V. Perceptual integration and perceptual segregation of brief visual stimuli. Vision Research, 1974, 14, 1056-1069. Liss, P. H., & Haith, M. M. The speed of visual processing in children and adults: Effects of backward and forward masking. Percepfion and Psychophysics, 1970, 8, 396-398. Mackworth, J. F. Some models of the reading process: Learners and skilled readers. Reading Research Quarterly, 1972, 7, 701-733. Miller, L. K. Visual masking and developmental differences in information processing. Child Development, Neisser, U. Cognitive
Spitz, H. H., &Thor, and
1972, 43, 704-709. psycho/ogy. New York:
Appleton-Century-Crofts, 1967. D. H. Visual backward masking in retardates and normals. Perception
Psychophysics,
1968, 4, 245-246.
Scheerer, E. Integration,
interruption and processing rate in visual backward masking. 1973, 36, 71-93. Stanley, G. The processing of digits by children with specific disability (Dyslexia). Brirish Psychologische
Journal
Forschung,
of Educational
Psychology,
1976, 46, 81-84.
Stanley, Cl., & Hall, R. Visual information processing in dyslexics. Chi/d Deve/opment, 1973, 44, 841-844. Taylor, M. M., & Creelman, C. D. PEST: Efficiency estimates on probability functions. The Journal of rhe Acoustical Sociery of America, l%7, 41, 782-787. Vellutino, F. R., Steger, J. A., Kaman, M., & De Setto, L. Visual form perception in deficient and normal readers as a function of age and orthographic-linguistic familiarity. Correx, 1975, 11, 22-30. Wald, A. Sequenria/ ana/ysis. New York: John Wiley & Sons, 1947. Welsandt, R. F., Zupnick, J. J., & Meyer, P. A. Age effects in backward visual masking (Crawford paradigm). Journal of Experimenral Child Psychology, 1973, 15, 454-464. RECEIVED: January 3, 1978; REVISED: March 21, 1978.