Myth #8: Scientific Measurement of Reading Speed
It is often thought that measuring reading speed is a simple matter of timing the reading interval while the examinee reads the test text, and then checking the examinee's comprehension is of the required standard. However, this does not account for the time spent reorganising information read, into something the reader understands. This is best exemplified by the period spent rereading the text in the mind's eye after the initial skim in the process of photo-reading. This mental rereading can and does occur when answering the comprehension questions, and thus depending on the development of short term memory, some examinees will exhibit higher reading speeds than they are truly capable of in conventional reading speed tests.
The answer is to time the comprehension test as well as the reading, but how do we separate differences of hand-eye coordination as they affect the time spent either typing or writing answers or otherwise selecting options from multiple choice questions or items in drop-down lists? Thus obtaining definitive reading speeds from single comprehension tests is as yet an unanswered question in terms of a repeatable and meaningful scientific method.
The closest one can get to a scientific test of reading speed at a set comprehension level is to employ the CooP system as follows:
- Display questions and text simultaneously.
- Ensure the word count includes the number of words for both text and questions.
- Each heading clause is counted as only a single word to filter out the highlight skimming effect.
- The test time begins with the appearance of the simultaneous text & question display.
- The test time ends when all questions are answered and the minimum comprehension standard is met.
- Separately time parts of equal grade but substantially different word count to filter out coordination lag.
- Multiple choice is used to test both structure and object recall.
(Answers are not written or typed because exact word testing only measures memory and written answers are just as prone to misinterpretation by examiners, depending on differences in spoken dialect between examiner and examinee.)
Multiple choice guesswork can be overcome by the following scoring system:
Maximum Score (100% correct) = Total number of choices - Total
number of questions
Actual Score = Maximum Score - Total number of choices in every incorrect answer
Statistically, multiple guessers will average 0 thereby rendering the mean advantage of guessing unknown or unviable answers negligible. The number of baseline questions generally will ensure that all but unlucky multiple guess practitioners will initially score above zero. This does not account for the time spent on hand-eye coordination activities such as manipulating the mouse or a pencil, but it does allow for for the inclusion of mental rereading as well as actual rereading (regression is part of the way some people read) in the time for the test.
Mathematically, by using a two part test in which both parts are of an equal reading grade and equal questionnaire word count, hand-eye coordination lag time in speed-reading tests is at least theoretically measurable. If the number of words attributed to Part B is double the number of words attributed to Part A, the amount of time by which Part A exceeds half the time taken for Part B is statistically (IE on average) equal to one third of the total hand-eye coordination lag time affecting the test.
In this sense, some statistical methods can be applied to reading speed tests but these methods are only reliable over a significant number of tests. Reading speed can be tested scientifically, but statistical methods need to be applied over a statistically significant number of tests for results to be repeatable (within margins of error). However, rigorous scientific methodology is not used in commercial reading speed tests because the amount of work required on the part of the client is generally enough to hurt product sales. Thus the assertion that reading speed measurements are more than an approximation is nearly always bogus.
