Here, we describe the construction of a test to formally evaluate sustained silent reading. We demonstrate that the sustained reading test has good test–retest reliability, particularly with regards to reading speed. Our data, from subjects with normal vision and VF loss from glaucoma, also demonstrate that out loud tests of reading do not accurately reflect sustained silent reading test ability for a number of reasons. First, the correlations between sustained silent and out loud reading speeds are lower than the correlation observed between two measures of out loud reading speed. Second, differences between sustained silent and out loud reading speeds vary proportionally with the magnitude of reading speed, demonstrate broad limits of agreement, and show extreme variability among rapid silent readers. Finally, measurement of sustained silent reading demonstrates the ability to capture novel aspects of reading not captured by out loud reading measures, such as changes in reading speed over time that may reflect reading fatigue. Many reading tasks involve sustained silent reading, and this novel test offers a new and useful tool for relating vision to the ability to effectively perform reading tasks.
The distinction between out loud and silent reading has been discussed previously, with Carver describing silent reading done for the purpose of comprehension as “rauding.”
17,19–21 In a group of patients with AMD, Lovie-Kitchen demonstrated a high correlation between silent and oral rauding rates (
r = 0.90), though silent reading was considerably faster.
19 This work lies in contrast to the current findings, which demonstrated a considerably poorer correlation between out loud and silent reading (
r = 0.59–0.68). Differences in our findings may have arisen from the fact that our subjects had normal or near-normal visual acuities as opposed to poor vision, read much faster than the Lovie-Kitchen cohort,
19 and much more frequently demonstrated silently read at rates at which they were likely incapable or willing to speak. Additionally, correlation coefficients have significant limitations as a measure of agreement,
16 and our Bland-Altman comparisons of sustained silent and out loud reading measures demonstrated evidence of proportional error (
Fig. 3), and extremely broad limits of agreement (Table). This proportional error could not be resolved by simply log-transforming both reading speeds,
18 likely because linear increases in out loud reading speed correspond to exponential increases in sustained silent reading speed. These findings suggest that short duration out loud reading speed does not adequately capture sustained silent reading speed in patients with normal visual acuity, and that directly testing silent reading is important, particularly when evaluating reading in individuals with normal sight and/or visual limitation not defined by decreased visual acuity. Further work is necessary to determine whether the observed discordance is solely due to the mode of reading (silent versus out loud), or is also a function of other test differences such as duration of testing, type of reading material, or presentation style of reading material.
Another feature of the sustained silent reading test presented here is that it is able to measure changes in reading speed over time (illustrated in
Fig. 4). Complaints detailing an inability to sustain reading over long periods of time are common in a clinical setting, and changes in reading speed may reflect such complaints. Indeed, as demonstrated in our companion paper, glaucoma was associated with a greater likelihood of decreasing reading speed over time than glaucoma suspects, which may at least partially explain the frequent clinical complaints of “fatigue” while reading in this population.
13 While the magnitude of reading speed declines were small, it is possible that they reflect significant fatigue that may lead patients to “put down the book” earlier than they would like.
Considerable effort was devoted to developing and validating comprehension questions corresponding to the silent reading passage. Comprehension questions were introduced to avoid having subjects simply skim the reading material in order to complete the passage faster, though the apparent value added from our testing of comprehension was not clear. We observed that faster readers tended to perform better on comprehension questions, suggesting that greater intelligence/cognitive ability allows both faster reading and better comprehension, even though for the same individual speed is likely to decrease as accuracy increases. This finding suggests that poor concordance between silent and out loud reading was not a result of fast silent readers paying less attention, thus, reflecting a speed–accuracy tradeoff previously described by Carver and others.
17,21 More likely, administration of comprehension questions minimized the variation in attention given to the reading material; thus, minimizing this source of reading speed variability. As comprehension could be calculated as a continuous variable, analytic models in the companion paper also incorporated this factor as a covariate.
13 The threat of comprehension questions may also have lead some individuals to read slower than what they were capable of, which may (along with reading fatigue) explain why several individuals demonstrated slower silent reading speeds than out loud reading speeds.
A significant obstacle to measuring silent reading is the time required for test administration. We administered our test for 30 minutes, the longest duration we felt was realistic without giving the subject a break. However, our findings suggest that accurate estimates for sustained silent reading speed and reading speed slope can also be obtained in nearly all patients with shorter testing durations (15–20 minutes). The accuracy of data obtained over these shorter durations also mitigates concerns that our data may be inaccurate in the few individuals who read so quickly that they completed the reading material well before the full 30 minute reading period had elapsed.
There are limitations to our sustained silent reading test, which may limit its utility. The text presented was only one size, and was relatively small print. Therefore, the test as constructed would not be useful in evaluating individuals with significant visual acuity loss from AMD or other conditions, nor was the test validated in such a low vision population. Additionally, there is a limited amount of reading material available, which limits the number of conditions under which reading may be tested. Testing is also time consuming and may not be feasible in a clinical setting or even in a research setting where time is limited. Reading speed is also determined by several nonvisual factors such as language development, producing a variability that we did not account for.
17,22,23 Finally, the text was written at the sixth grade level, and may not be appropriate for individuals who do not have at least some high school education.
Despite these limitations, to our knowledge, there are no standardized tests available to evaluate silent reading speed and change in reading speed over long durations of time. Given that sustained silent reading is a central feature of many types of employment and is also a requisite for many leisure activities, evaluating reading under such conditions may provide new insights into the degree of disability experienced among patients with vision loss and/or the extent to which visual rehabilitation is successful. In particular, our sustained reading test would be useful in evaluating how reading is impacted when visual acuity is normal or nearly normal, and in conditions where the inability to sustain reading is a frequent complaint, such as dry eye, computer vision syndrome, and glaucoma.