Spatial Frequency Sensitivity Differences between Adults of Good and Poor Reading Ability

Abstract

purpose. To determine whether normal adults of good and poor reading ability exhibit different patterns of sensitivity to spatial frequency, as previously found between dyslexic and nondyslexic control subjects.

methods. The visual acuity, spatial frequency sensitivity, and reading ability of 96 normal, nondyslexic adults was assessed. Participants were ranked according to reading ability. The top 50% were classified as good readers and the bottom 50% as poor readers.

results. Despite no differences in visual acuity, good and poor readers showed different patterns of spatial frequency sensitivity. In particular, compared with good readers, poor readers showed reduced sensitivity to spatial frequencies between 2 and 6 cyc/deg, and no differences in sensitivity were found at lower or higher spatial frequencies.

conclusions. The findings indicate that spatial frequency sensitivity differences found previously between dyslexic and nondyslexic controls can extend to the normal (nondyslexic) adult population.

Ophthalmic clinicians have long used letter charts as the standard measure of vision necessary for fluent reading. However, letter charts measure a person’s ability to resolve fine detail and, even when visual acuity is normal, other visual factors may also play an important role in reading. Spatial frequency sensitivity is one such factor of increasing interest. 

Spatial frequency sensitivity provides an indication of a person’s ability to perceive visual information across the full visual spectrum, from fine to broad scale, and is measured with repetitive patterns of black-and-white bars. The luminance pattern across these bars describes a sine wave, and the patterns are referred to as sine-wave gratings. Spatial frequency is expressed as the number of cycles (one black plus one white bar) per degree of visual angle. The amount of contrast required (i.e., contrast threshold) for detection of sine-wave gratings is known to vary systematically as a function of their spatial frequency. The reciprocal of the threshold contrast (i.e., 1/contrast threshold) is the contrast sensitivity, and the contrast sensitivity function (CSF) describes the variation of the sensitivity over a range of spatial frequencies. ^{1 2 3}  

A substantial body of evidence indicates that dyslexic and nondyslexic control subjects exhibit different patterns of sensitivity to spatial frequencies. ^{4 5 6 7 8 9 10 11 12 13 14} In particular, study results indicate that dyslexic individuals have reduced sensitivity to certain spatial frequencies, and this reduction tends to be greatest in the low- to midfrequency range (i.e., between 2 and 8 cyc/deg). ^{4 5 6 7 8 10 11 13 14 15} However, despite a wealth of research involving dyslexic individuals there is a dearth of studies examining spatial frequency sensitivity differences between “normal,” nondyslexic adults of good or poor reading ability. 

Surveys of reading ability reveal a considerable range of reading abilities in the normal adult population. ^{16 17} For instance, the International Adult Literacy Survey (IALS; 1994–1998, National Literacy Secretariat, Canada) ¹⁷ indicates that up to 52% of the normal adult population, although able to read, have sufficiently low levels of reading ability to make it difficult for them to face novel demands, such as acquiring new work skills. Accordingly, the purpose of the present study was to assess the reading ability of normal adults and determine whether good and poor readers exhibit different patterns of sensitivity to spatial frequencies. 

The present study drew on a sample of normal, nondyslexic adults recruited from the university population. Reading ability was assessed by measuring effective reading speed. Effective reading speed is a measure widely used to assess the reading ability of normal (nondyslexic) adults. ^{18 19 20 21 22 23 24 25 26 27 28} It is defined as the time taken to read short passages or sentences in words per minute (wpm) multiplied by either the number of words read correctly ²⁶ or the number of questions answered correctly in a brief comprehension test. ¹⁹ As a combined measure of reading speed and comprehension, effective reading speed encapsulates not only participants’ ability to comprehend what they have read but also the efficiency with which they are able to achieve understanding, and this combined measure of speed and comprehension was selected as the index of reading ability in our study. 

Contrast sensitivity was measured with a spatial two-alternative, forced-choice (2AFC) task in which participants were required to indicate on which side of a display screen a sinusoidal grating was presented. The QUEST staircase procedure was used to estimate each participant’s contrast threshold. ²⁹ This procedure has been effective in several investigations ^{26 30} and enabled assessment of participants’ sensitivity to a range of spatial frequencies from 0.5 to 12 cyc/deg. Comparison of sensitivity to spatial frequencies between good and poor readers promised to reveal whether normal adults of good or poor reading ability differ in their sensitivity to certain spatial frequencies. 

Method

Participants

Ninety-six undergraduate students (38 men and 58 women) between the ages of 18 and 35 took part in the experiment. All participants were native speakers of English, and none reported any history of epilepsy or dyslexia or demonstrated any reading problems when tested. Each participant took part in one 75-minute session, in which each participant was tested for visual acuity, contrast sensitivity, and reading ability. Informed consent was obtained from each participant before the experiment, in accordance with the Declaration of Helsinki. 

Visual Acuity

Visual acuity was tested with the Bailey-Lovie eye chart. ³¹ Participants were required to continue reading letters down the chart from a distance of 3 m until they failed to identify any letters on one line. Performance was scored by the method recommended by Kitchin and Bailey. ³² The total number of letters read incorrectly was recorded, and an “error” score of 0.02 assigned to each. These scores were added to the last line on which any letters were read. To continue participating in the study, participants were required to have a minimum binocular acuity of 10/10 (3/3), indicative of normal visual acuity. 

Contrast Sensitivity

Stimuli.

Contrast sensitivity was tested with gray-scale vertical sine-wave gratings of 0.5, 1, 2, 4, 6, 8, 10, and 12 cyc/deg. These spatial frequencies were chosen to conform to previous psychophysical studies of spatial frequency sensitivity ^{1 2 3} and to cover the range of spatial frequencies used in previous studies involving dyslexic individuals. ^{4 5 6 7 8 9 10 11 12 13 15} A Gaussian patch modulated each sine wave grating, to create eight Gabor stimuli, each with a different spatial frequency. 

Visual Conditions.

Viewing was binocular. The Gabor stimuli were presented on a γ-corrected video monitor with a resolution of 980 × 1024 pixels. Viewed from a distance of 57 cm, the viewable area of the monitor measured 23° horizontally and 29° vertically. Background illumination of the monitor screen and space-averaged luminance of each Gabor was kept constant at 35 cd/m². Each Gabor subtended 12° vertically and 12° horizontally (the radial size of each standard deviation of Gaussian patch was 3°) and were presented so that the center of each Gabor always fell 6° to the left or right of the center of the video monitor on the horizontal midline. 

Apparatus.

The Gabor stimuli were presented on a 40.4 × 30.2-cm monitor (Trinitron GDM-F520; Sony, Tokyo, Japan). A Cambridge Research Systems (Rochester, UK) visual stimulus generator (VSG2/5) card controlled stimulus presentations and timing. Responses were collected with a button box (CT3; Cambridge Research Systems). Luminance was measured with an optical photometer. The experiment was conducted in a quiet, darkened room. A viewing hood fixed to the monitor ensured a constant viewing distance and eliminated any extraneous light sources. 

Design.

Each different Gabor stimulus was presented 80 times, randomly interleaved, giving a total of 640 trials. Contrast sensitivity was measured with a spatial 2AFC task in which participants had to decide on which side of the video monitor the Gabor patch was presented. During each trial, the contrast of each Gabor was determined by using the QUEST algorithm ^{29 30 33} in the Psychophysics Toolbox. ^{34 35} The threshold was set at 0.82, and the initial contrast of each Gabor was set at the average obtained from pilot studies. The final estimate was taken as the mean of the posterior probability distribution function. ³⁰  

Procedure.

Each participant was given written instructions informing him or her of the task and of the importance of responding as accurately as possible. At the start of each trial, a clearly audible “beep” was emitted from the button box to inform participants that a Gabor patch was about to be presented. A Gabor patch was then presented on either the left or right side of the video monitor. To avoid onset transients, each Gabor was ramped on (exponentially) over the first 100 ms. Each Gabor then remained on the video monitor at the contrast level determined by the QUEST algorithm, until a choice response was made. To make a choice, participants were required to press one of two buttons to indicate on which side of the video monitor the Gabor patch was presented. 

Reading Speed

Stimuli.

Seven passages were selected from Notes From a Small Island by Bill Bryson, ³⁶ which provided an engaging text. On average, each passage contained 527 words. After each passage, five multiple-choice questions were presented. The questions referred to different detailed aspects of the preceding paragraph and were designed to ensure that participants read each paragraph in full. ^{37 38}  

Visual Conditions.

Viewing was binocular. Each passage was presented on the same γ-corrected video monitor as that used to test contrast sensitivity. The text was presented in black on a light gray background, and a complete passage of text filled an area approximately 18° (horizontal) × 27° (vertical) and had proportions similar to those of an A4 page of text (which is familiar in the British reading environment). Background illumination of the monitor screen was 46 cd/m², and the luminance of text was 0.15 cd/m². 

Design.

Each participant was presented with all seven passages. One passage (the first shown) was always used as practice, and the remaining six as test passages, shown in a random order. 

Procedure.

Participants were told that the experiment would examine the time taken to read different passages of text, and that they should read through each passage once, from start to finish, as rapidly as if they were reading a page of a book. As soon as a button was pressed, a passage was presented (shown in its entirety on the screen) and the timer started. Participants pressed the button again when they had read the final word of each passage, and this stopped the timer. The passage was immediately replaced with five multiple-choice questions and participants were required to select one of three answers from each of the five questions before continuing. 

Results

To identify good and poor readers, an effective reading speed was calculated for each participant by multiplying the reading speed in words per minute by the proportion of questions they answered correctly. Effective reading speed ranged from 99 (reading speed, 156 wpm; proportion of questions answered correctly, 0.63) to 365 (reading speed, 421 wpm; proportion of questions answered correctly, 0.87). For poor readers (bottom 50%, comprising 20 men and 28 women), effective reading speeds ranged from 99 to 185 (mean, 53 ± 21 wpm; SD) and, for good readers (top 50%, comprising 18 men and 30 women), from 186 to 365 (mean, 228 ± 37 wpm). 

Visual acuities for all participants ranged from −0.5 to −0.3 logMAR (mean, −0.35 ± 0.05; SD). The visual acuity data were analyzed with an analysis of variance with two between-subjects factors (reading ability, sex). No statistically reliable differences in visual acuity were found between good and poor readers (F_(1,92) = 0.07; P > 0.70), between men and women F_(1,92) = 1.75; P > 0.10), and no interaction was found between reading ability and sex (F_(1,92) = 0.15; P > 0.70). Nevertheless, good and poor readers exhibited different patterns of spatial frequency sensitivity. The results of the spatial frequency sensitivity test for good and poor readers are shown in Figure 1 . The sensitivity data were analyzed with a Greenhouse-Geisser corrected analysis of variance with two between-subjects factors (reading ability, sex) and one within-subject factor (spatial frequency). This analysis revealed main effects of reading ability (F_(1,92) = 6.00; P < 0.05) and spatial frequency (F_{(3.97,365.60)} = 310.06; P < 0.001) and an interaction between reading ability and spatial frequency (F_{(3.97,365.60)} = 3.05, P < 0.01). No differences were found between men and women and no interactions were obtained with this factor. 

The Tukey honest significant difference (HSD) tests showed that poor readers were less sensitive than good readers to spatial frequencies of 2, 4, and 6 cyc/deg (all P < 0.05). In good readers, sensitivity was lower at spatial frequencies of 1 and 4 cyc/deg than for 2 cyc/deg (P < 0.01); lower for 0.5 and 6 cyc/deg than for 1, 2, and 4 cyc/deg (P < 0.01); lower for 8 cyc/deg than for 0.5, 1, 2, 4, and 6 cyc/deg (P < 0.01); and lower for 10 and 12 cyc/deg than for 0.5, 1, 2, 4, 6, and 8 cyc/deg (P < 0.05). However, for poor readers, sensitivity was lower at spatial frequencies of 1 cyc/deg than of 2 cyc/deg (P < 0.01); lower at 4 cyc/deg than at 1 and 2 cyc/deg (P < 0.01); lower at 0.5 cyc/deg than at 1, 2, and 4 cyc/deg (P < 0.01); lower at 6 cyc/deg than at 1, 2, and 4 cyc/deg (P < 0.01); lower at 8 cyc/deg than for 0.5, 1, 2, 4, and 6 cyc/deg (P < 0.01); and lower at 10 and 12 cyc/deg than at 0.5, 1, 2, 4, 6, and 8 cyc/deg (P < 0.01). No other comparisons were significant. 

Statistically, a similar pattern of results emerged when reading ability was considered in terms of raw reading speed alone. However, effective reading speed registers differences in raw reading speed and comprehension combined. Indeed, this point is further supported by the high correlation between effective and raw reading speeds (r = 0.90; P < 0.001) and between effective reading speed and comprehension (r = 0.34; P < 0.001). Therefore, as in previous studies, effective reading speed is considered the most appropriate measure of reading ability (see Jackson and McClelland ¹⁹ for further discussion). Further analysis revealed no speed accuracy tradeoff, as evidenced by the low correlation between raw reading speed and comprehension (r = −0.08; P > 0.40). 

Further correlation analysis was conducted to examine relations between effective reading speed and spatial frequency sensitivity. Figure 2shows the scatter plots relating effective reading speed and sensitivity to each spatial frequency separately for each spatial frequency. Statistically significant correlations were found between effective reading speed and sensitivity to spatial frequencies of 2 (r = 0.20; P < 0.05), 4 (r = 0.30; P < 0.01), 6 (r = 0.30; P < 0.01), and 8 (r = 0.27; P < 0.01) cyc/deg. No other correlations were significant. No statistically reliable relationship was found between visual acuity and effective reading speed (r = −0.04; P > 0.70). 

Discussion

The results of the present study indicate that sensitivity to certain spatial frequencies affects reading ability in the normal (nondyslexic) adult population. In particular, despite having normal visual acuity, poor readers showed reduced sensitivity to spatial frequencies of 2, 4, and 6 cyc/deg, relative to good readers. This finding resonates with those found previously between dyslexic and nondyslexic control subjects, ^{4 5 6 7 8 10 11 13 14 15} indicating that differences in spatial frequency sensitivity between dyslexic and nondyslexic control subjects can extend to the normal (nondyslexic) adult population. 

Further examination of the relationship between spatial frequency sensitivity and reading ability revealed moderate correlations between effective reading speed and spatial frequency sensitivity for spatial frequencies of 2, 4, 6, and 8 cyc/deg. Similarly, prereaders’ sensitivity to spatial frequencies of 2 and 4 cyc/deg has been found to be related to their reading ability 2 years later. ³⁹ The implication of the results in that study is that spatial frequency sensitivity differences found between children do not result from a failure to learn to read, but that differences in sensitivity to certain spatial frequencies produce different reading abilities and, in light of the present study, that relations between spatial frequency sensitivity and reading ability found previously with young children can extend into adulthood. 

However, results of previous studies that have tested performance with a small range of spatial frequencies (typically, no more than four) and have shown a decline in sensitivity to all spatial frequencies tested may reflect merely an overall lack of attention to the task ⁴⁰ or, in the case of normal, nondyslexic children, overall differences in intellectual ability. In contrast, the present study tested sensitivity over a wider range of spatial frequencies (eight, from 0.5 to 12 cyc/deg), and revealed no overall deficits for poor readers but, instead, deficits only at specific spatial frequencies. Consequently, it seems unlikely that problems concerning overall differences in intellectual ability between good and poor readers, or overall differences in attention to the task, can account for the selective reduction in spatial-frequency sensitivity displayed by adults of poor reading ability. 

The deficits displayed by poor readers in sensitivity to some, but not all, spatial frequencies provide a strong indication that differences in sensitivity to certain spatial frequencies affect the ability to read. However, the precise nature of the causal relationship between spatial frequency sensitivity and reading ability has yet to be fully determined. Indeed, although it currently seems highly unlikely that differences in reading ability cause selective differences in spatial frequency sensitivity, further rigorous experimental research is needed, to determine fully the precise link between spatial frequency sensitivity and reading ability. 

One explanation of the present findings is that poor readers have reduced sensitivity to a midrange of spatial frequencies, and this reduction in sensitivity to these spatial frequencies affects their ability to perceive words. Psychophysically, words are complex images comprising a broad range of spatial frequency information, from coarse to fine scale, and there is a growing body of research that suggests an important role for coarse-scale (low-frequency) and fine-scale (high-frequency) visual information in word perception ^{20 22 41 42 43 44 45} (Boden C, et al. IOVS 2000;41:ARVO Abstract 2298; Dakin SC, et al. IOVS 1999;40:ARVO Abstract 184). Indeed, several investigators posit a system of word recognition in which both coarse-scale visual information about word shape and more fine-scale information about letter features play an important role in visual word perception. ^{46 47 48 49} Consequently, individual differences in sensitivity to certain spatial frequencies may affect the weighting attached to different spatial frequencies in word perception, and people with different patterns of sensitivity to various spatial frequencies may rely on different spatial frequencies for recognizing words. 

Further research is needed to determine the precise role of different spatial frequencies in word perception, but spatial frequency sensitivity is emerging as a useful measure of the functional “quality” of vision. Accordingly, continued investigation of the relationship between spatial frequency sensitivity and reading ability promises to provide a clinically useful guide to the visual capabilities of normal adults. The findings of the present study, in conjunction with those obtained with dyslexic individuals ⁴ , ^{5 6 7 8 10 11 13 14 15} and normal, nondyslexic children ³⁹ underscore the importance of assessing reading ability from an ophthalmic perspective, as well as from the perspective of higher order mechanisms. Indeed, our findings show that it is likely that differences in patterns of sensitivity to spatial frequency play a major role in the individual reading ability of adults. ^{39 47}  

 Supported by Wellcome Trust Grant 059727 (TRJ).

 Submitted for publication November 15, 2003; revised July 4, 2004; accepted August 5, 2004.

 Disclosure: G.R. Patching, None; T.R. Jordan, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Log sensitivity (1/contrast threshold) for good and poor readers.

Figure 2.

 

The relationship between effective reading speed and spatial frequency sensitivity separately for each spatial frequency. The correlation coefficient, Pearson’s r, is specified in each scatterplot.

The authors thank Sharon Thomas for useful comments regarding various aspects of this work