An interesting side light is thrown on M.'s memory by his work in another experiment in which he was a subject. This experiment required that the subject look at an object for 10 secs. and then after the disappearance of its after-image manipulate the memory image. M. showed unusually persistent after-images. The memory images which followed were unusually clear in details and also persistent. They were moreover retained for weeks, as was shown by his surprising ability to recall the details of an image long past, and separated from the present one by many subsequent images. His memory was capacious rather than selective. His eyesight was tested and found to be normal for the range of the apparatus. Possibly his age (55 yrs.) is significant, although one of the two subjects who showed the greatest preference for objects and movements, Mo., was only six yrs. younger. The ages of the other subjects were S. 36 yrs., Hu. 23 yrs., B. 25 yrs., Ho. 27 yrs. That some if not all of the subjects did not have objective images in many of the noun and verb couplets if they were left to their own initiative to obtain them is evident from the image records in the _A_ set, in which the presence of the objective images was optional but the record obligatory. The same subject might have in one noun or verb series no visual images and in another he might have one for every couplet of the series. After the completion of the _A_ set, the effect of the presence of the objective images in series of 10 nouns alone, or 10 objects alone after two days' interval, was tested. This was merely a repetition of similar work by Kirkpatrick after three days' interval, and yielded similar results. As a matter of fact some of the subjects were unable wholly to exclude the objective images, but were compelled to admit and then suppress them as far as possible, so that it is really a question of degree of prominence and duration of the images. The presence of the objective images having been shown to be an aid in the case of series of nouns, the subjects were henceforth requested to obtain them in the noun and verb series of the _B_ and _C_ sets, and the image records show that they were entirely successful in doing so. 2. The total number of couplets in any one or in several sets may be divided into two classes: (1) Those in which indirect associations did not occur in the learning, and (2) those in which they did occur. For reasons already named we may call the first pure material and the second mixed. We can then ascertain in each the proportion of correctly recalled couplets after one, two, nine and sixteen days' interval, and thus see the importance of indirect associations as a factor in recall. This is what has been done in the following table.