of the eye marked near the pupil with a flake of Chinese white.1 Against the method here employed they offer a criticism. They maintain that, as the cornea is not flat, it is too imperfect a reflector to make the records reliable. The movements recorded are said not to correspond in amount to those that the eye makes. Professor Dodge on the other hand has comparable records, as yet unpublished, taken by both methods which go to show that the criticism is much exaggerated, if not entirely invalid. At any rate, the criticism, even if true, would not affect my results. The only data I used are the number of pauses per line, the faithful record of which is not questioned.

In addition to getting records with a camera, I got some by means of a mirror, the method used by Erdmann and Dodge.2 In the application of this method the reader and the experimenter are seated on opposite sides of a narrow table. The reading matter is laid flat upon the table and beside it is placed a mirror. The reader takes a comfortable position, supporting his head with both hands so as to minimize head movements. The operator gets a good reflection of the reader's eyes in the mirror and after some practise can readily observe, count and record all movements. At first the task of catching all the movements seems hopeless, but practise will soon give one confidence in his count. It is easier to observe the movements than the pauses, but when one does this he should remember, when coupling his results, that there is always one more pause per line than there are movements.

In four cases the mirror records duplicate the photographic records, and in two they stand alone.

A glance at Table XVI shows that the results obtained by the two methods are in close agreement, indicating that the mirror method is accurate enough for most purposes when one wants to obtain only the number of pauses per line. The discrepancy is largest in the records of subject H. R. She is quite a rapid reader but makes many eye movements per line, most of which are small and so hard to count.

The passages read in this connection were taken from Palmer's lecture on The Glory of the Imperfect and from the educational journal referred to above. They were placed approximately 30 cm. from the eye, the distance used in mapping the field of vision. The same lines were not always read by the different subjects. Skipping about was necessary because, after several trials with the same passage, it became too familiar to give trustworthy results. A comparison of

1 Psychological Review, Monograph Supplement, Vol. VII, No. 1.

2 Op. Cit., pp. 46–47.

the records from the same subject showed, however, that the number of pauses per line did not materially vary in different selections from the same article.

The lines in 'Palmer' are 80 mm. long and average 9 words per line, while those in 'education' are 125.5 mm. long and average 14 words per line. In the average number of words read per fixation, the subjects varied from 1.2 to 2.5 in 'Palmer' and from 1.4 to 3.2 in 'education.' The average number of words read per fixation was 1.6 for 'Palmer' and 1.7 for 'education.' The average extent of the visual field covered per fixation varied from 11 mm. in the slowest reader to 31 mm. in the fastest. In all of these subjects the horizontal extent of the field within which they could still distinguish the n and the u 90 per cent of the time varied from 40 to 71 mm., or from 7.6 to 12.5 degrees. It was 40 mm. in the slowest reader and 43 mm. in the fastest. This indicates that probably not even the most rapid reader ever made use of the entire extent of his visual field available for seeing words. There is consequently no time lost in reading, owing to the way the words are arranged on the page. With print of ordinary size, we can take in the material faster than the mind is able to assimilate it, and there is no call for methods of arranging the words so that more can be crowded simultaneously into the field of distinct vision.1

1 Cf. Cattell, Phil. Stud., Vol. III, p. 125.

Palmer.

M. V.

Education.

M. V.

The correlations with reading rate are given in Table XVII. The only pronounced correlation is found inversely with the number of reading pauses. The rapid readers make quite uniformly fewer TABLE XVII.-CORRELATIONS WITH READING RATE, ETC.

pauses than the slow readers. The correlation is so high that a vital connection between the two phenomena cannot be doubted. In twelve cases the percentage of displacement is 16 and the Pearson coefficient +.81.

Dearborn did not work out any coefficients of correlation between reading rate and reading pauses and his data are not so given that it is possible for us to do so, but he reaches substantially the same conclusion as I do.' He says (p. 122): "A wider 'spanning' of attention-as denoted both by the greater frequency of long pauses at the beginning of the line and by fewer fixations per line-is characteristic of the more rapid readers. The slow readers have a narrower span, or working extent of attention, and a greater total arc of movement."

The Pearson coefficient of correlation between the horizontal extent of acute vision and reading rate is .06 and the percent of displacement 56. Correlated inversely with the number of reading pauses the corresponding figures are -.10 and 42 percent. The fact that there is no correlation between these phenomena came somewhat as a surprise. One would think that a larger area would enable a person to see distinctly a larger portion of a line and so allow him to make fewer fixations and read more rapidly. But such is 1 Op. Cit., Chapter XIII.

evidently not the case. Visual perception seems to be in a large measure independent of the quality of the eyes. The percent of displacement between reading rate and visual acuity is 35. Arranging the eyes again in their order of freedom from defect, the percentage of displacement with reading rate is 45.

The coefficients of correlation of other tests of visual perception show a similar independence of the quality of the eyes, as well as a lack of correlation among themselves.

The "A" test consists of marking as rapidly as possible one hundred A's distributed by chance among four hundred other capital letters. The time is taken with a stop-watch. It is supposed that the test measures rate of perception. Most of the subjects marked the sheet four different times, but with five circumstances permitted it to be done only once.

The results, which are summarized in Table XVIII, are given in two columns. In the first column is given the average time in secTABLE XVIII.—'A' TEST.

onds the subject took to see and mark one A, while in the second column is given the time with the marking or motor time subtracted. In connection with the A test each subject marked as rapidly as possible one hundred dots in a similar manner as he marked the A's, the time thus consumed was taken to be the subject's motor time, and was subtracted from the results as given in the first column. Both results are given because it is not clear that this time should be subtracted. The time it takes to mark the A's is not all lost from perception; in fact, only a small fraction of it can be regarded as lost. A person does not not look at the A till he has marked it. As soon as the letter is recognized, the eye passes on to look for another while the hand in a semi-automatic way marks it. Furthermore, the eye seldom, if ever, has to wait for the hand to catch up. It is kept continuously busy looking for letters.

The correlation between the horizontal extent of acute vision and the A test is 42 percent of displacement when the motor time is

not subtracted, and 50 percent when it is. This indicates very little correlation, if any. One might again have supposed that a larger area would be of assistance in finding the letters, but it apparently is not.

When the rank in reading rate is compared with that in the A test, the displacement is 33 per cent; or, when the motor time is subtracted, 40 per cent. The Pearson coefficient in the second instance is +.36. This indicates some correlation but not a great amount. Quantz1 found a higher correlation between reading rate and perceptive power. He represents the correlation diagrammatically and does not give a numeral coefficient, but the meager data he supplies give a Pearson coefficient of +.88.

The discrepancy between his conclusion and mine no doubt results from the different measures of perception used. Perception is too complex a process to be adequately measured by the A test. Quantz measured it by means of geometrical figures, colors, isolated words, and sentences. These were exposed for a short time (′′ and 1"), and "the subject was required to name aloud, in order, and as rapidly as possible..... as many as he was able to see while the card was exposed" (p. 2). The card was shown until everything on it was read, the subject beginning to read at each exposure where he had ended in the preceding. It is at once apparent that these tests were far more analogous to the reading process than the A test. In fact, the tests with the words and the sentences were essentially reading, and the forms and colors were perceived much as words are perceived in reading. Instead of saying that these tests show a correlation between visual perception and reading, one is tempted to say that they show little more than a correlation between tachistoscopic and ordinary reading.

Quantz draws the conclusion also that these tests show that the reading process depends largely on physiological influences because he assumes that perception is largely a physiological function (p. 50). This is far from clear, for such a conclusion can no more be drawn from the tests used than from the reading process directly.

Another test that was made and correlated with the horizontal extent of acute vision and with reading rate was the number of small vertical lines that could be seen simultaneously during an exposure time of 50 sigma. The lines consisted of the capital letter I printed on cards with a Fay-Sholes typewriter. They were approximately 2.5 mm. apart.

A similar test was made by Cattell for the purpose of measuring 1 Psych. Review, Monograph Supplement, No. V, pp. 16-17.