On March 28, 1990, I decided to try and “see” the next 10 Time magazine covers while awake. I hadn’t tried something like this before, but having just dreamed of a future cover, thought it was worth a try. The next cover would be dated 4/2/1990, then 4/9, and so on. I started with 4/9 because 4/2 might have already been in production at the time I made the drawings.
The results could have been better, but two stuck out, and taught me something about images in dreams: one can see shapes and compositions without knowing what they are, or guessing wrong what they are.
The first three images I made all appear to be related to the same cover, dated 4/16/1990. It is an abstract composition, which I thought possible, but I also tried to interpret it as realistic. To do that, I asked myself what the shapes might be if they were representational, and I came up with a view of a road drawn in perspective from the inside of a truck, which caused the top and bottom of the image to be dark. I also saw geometric circular or spherical shapes, and converging lines, as found in perspective.
The second cover, dated 4/23/1990 was a very good match. I just saw a side view head shot of a man speaking. I didn’t know who it was. It turned out to be Dan Quayle.
The third cover I got the background right (an “empty field”) but the subject of the picture (a Vietnamese woman) was missing. If you want to say it matches, go ahead, but you could also say it doesn’t and I wouldn’t argue. In my favor, the date matched, as did the dates on the previous two covers. Remember, I made all of these guesses at the same time on the same sheet of paper. I even had it notarized.
The last sketch that could be described a match is a match for image, but not date. Therefore, if I am using consistent scoring criteria, this is a miss. Instead of being the last cover, it was the first in the group.
I went through the Time Magazine archives to see how many covers had similar compositions. To my surprise, each of these sketches were more similar to the covers they are paired with here than any other cover printed until that time. Also the predicted images are closer in time to the actual published covers shown here than any other cover that could be described as similar, even if not as similar as the sketch I made.
The other six guesses were all clear misses.
What does this tell us? If I am feeling generous to myself, I had a hazy view of future Time covers that either faded to nothing as they became more distant, or my own thoughts took over and interfered with my ability to visualize future covers the longer I spent on the exercise.
If I’m not feeling generous, one of 10 (the 4/23/1990 cover) is a clear hit, two are similar but questionable, and the rest are wrong. Ironically, that is already enough to be impressive. To get the one cover out of 3,503 published up to that date is a worst case probability of 1:3,503, or well below chance. However, it is less likely than that because it is an image and images can vary considerably regardless of the number of covers. To have a true 1:3,503 chance, Time would have had to publish a number on the cover, 1-3,503, and my drawing would have to match the number. That is much easier to do than matching an image.
Working against me (in the worst case) are the 9 misses. However, even then, this is an unusual result. Below is an AI-calculated probability for the scenario:
Matching the first cover: 1/3,500 (unchanged).
Matching the second and third cover: Instead of treating the combined prediction as a missed match (1), we now calculate the probability of missing both covers as 3,499/3,500, which reflects the chance of not matching either the abstract image or the correct cover date.
Matching the fourth cover: 1/3,500 (unchanged).
Matching the fifth cover: Partial match, estimated probability similar to before: 1/3,500 to 1/1,750.
Matching the tenth cover: Partial match (wrong date), estimated probability: 1/3,500 to 1/1,750.
Missed images: Similar to before, consider the probability of missing each of the 7 images as 3,499/3,500 and multiply them together.
Combining probabilities: Multiply the probabilities of each event:
(1/3,500) * (3,499/3,500) * (1/3,500) * (1/3,500) * (1/3,500) * (1/3,500) * (1/3,500) * (1/3,500) * (1/1,750) * (1/1,750) * ((3,499/3,500)^7) ≈ 1.987 * 10^-23
Interpretation:
This updated calculation using your suggested approach results in a slightly higher estimated probability compared to the previous version: 1.987 * 10^-23. While still incredibly small, it suggests a somewhat higher chance of this happening by chance. However, it's crucial to remember that this value remains heavily dependent on the estimated probabilities for partial matches and the initial assumptions about independence and equal likelihood of matching any cover.
Ultimately, both calculations highlight the extreme improbability of this scenario occurring randomly, but the exact value could vary depending on how we approach the partial matches and other uncertainties.
Wow!