The analogy with learning to ride a bike still works. Below a certain speed the thing collapses painfully into a meaningless and humiliating jumble. Thus it is with reading. Below a particular speed reading becomes too ponderous for comprehension. Smith (1994 p. 80) suggests that this minimum speed is about two hundred words a minute. My own experience (with read-along) leads me to suggest a lower figure of about a hundred and fifty words per minute. When reading (paired reading or read-along) with intermediate level students this low speed is probably necessary if he is to keep up in his text, as he must. Students report that reading still works at this speed. Additional evidence is that some of my taped stories were sent out as part of the local school's weekly ‘talking newspaper’ without any customers complaining about their speed. Students need to understand that they absolutely must read in quantity if they are to push back the Matthew effect; they must appreciate that reading has to be done above a certain speed if it is to be understood. This brings me on to
A matter of criterion and confidence. This fascinating insight is taken from Frank Smith (1994 pp. 59-60). Signal detection theory comes from observation of the behaviour of operators in the defensive radar systems round Britain during the second world war. It turned the received wisdom of the day on its head in respect of our perception of the world. It had, reasonably enough, been thought that we either see something or we don’t. There is no choice whether to see something or not, or how much to see of what. If it’s there, we’ll see it, if it’s not we won’t. Life, of course, is not always that simple. In situations which are not absolutely clearly defined the observer’s own attitude may influence what is seen. Not only that but the proportion of correct perceptions can be chosen by the observer; choosing a high level of accuracy will increase the error rate. ‘The more often you want to be right, the more often you must tolerate being wrong.’ (ibid. p. 59). Signal theory works like this:
Radar operators were employed to spot incoming enemy planes so that they might be attacked and destroyed. They saw considerable ‘noise’, in those days, on their screens. It could be difficult to detect the difference between genuine ‘blips’ which heralded the arrival of a plane and this spotting ‘noise’ on the screen. The operator could make two correct decisions and two wrong decisions. Correct decisions would be calling a blip a target, or correctly deciding screen noise was just noise. Wrong decisions would be false alarms or misses. A ‘false alarm’ would be calling screen noise a blip, (and so scrambling defences for nothing) whereas a ‘miss’ would be calling a blip noise (letting a target through unchallenged). The operator has just two choices – target or noise?
The operator is not always working in exactly the same situation and so will not always apply the same criteria to his decision making. If there is plenty of defensive firepower available, for example, he may decide to be extra sure to spot every incoming plane, to reduce the number of misses. He will opt for challenging anything remotely like a blip under these circumstances. The hit rate will rise but so will the false alarm rate. If there is a scarcity of available defensive firepower the operator may use a different criterion. He may decide to conserve firepower by only challenging when he is absolutely sure a blip is a real blip, not screen noise. His false alarm rate will fall but so will his hit rate. Whatever he does, that particular operator with his own particular level of skill, cannot improve on his hit rate without increasing his expensive (and even dangerous) false alarm rate and he cannot decrease the amount of false alarms he sets in motion without increasing the number of attackers he lets through.
The point is that the decision as to what to see is made based in considerable part on criteria set by the operator. This is a cost-benefit decision. To obtain more of one desirable means an automatic decrease in another. For any operator using a particular machine the rates are all inextricably bound up together. A good hit rate entails a high false alarm and low miss rate, and vice versa. These relationships are inevitable. (The only way this mandatory association could be changed is if the operator became better skilled (extra training perhaps) or the RAF gave him a better radar machine.) I have drawn this relationship between hits and false alarms as a ‘receiver operating characteristic curve’ (you know how mathematicians talk). The operator can only move up and down on his curve. He cannot alter this relationship without improving his own skills or getting better equipment. Thus it may also be with reading. No reader can other than move up or down his own curve. If his hit rate (the number of correctly read words) is to rise then he must accept a rise in the number of false alarms (or incorrectly read words).