Most folks know that people sometimes see pictures of things in actually random data, whether it’s faces in the clouds or Mother Theresa on a sticky bun. Michael Shermer did a nice video that talks about this on the TED web site. I want to focus on one example from Shermer’s talk, the “Face on Mars”. Here’s the famous image:
It looks like a face because there’s not much detail in the image. A higher resolution picture of the same area (click image for very high resolution version) of Mars doesn’t look like a face:
However, if you squint, you can may be able to see the face; that’s because when you squint, you introduce noise.
We see the same effect on the infamous “smiley face” on Alvin Onaka’s signature on Obama’s long form. In this case noise was introduced by the PDF optimization that removed information from the image by converting gray scale to black and white. First the noisy version:
Now the gray-scale version:
Basically this same noise trick appears in one form or another in most of the birther analyses of the Obama long form. Let me give you one final noisy example. This is from Paul Irey’s comparison of noisy (in this case pixelated) data from the long form in an attempt to argue that the font is different on different typed information:
What we actually see is noise from converting gray scale to black and white and changing resolution. Now, let me show you font changes from an actual fake birth certificate:
and a real variable and monospaced contradiction.
In these examples, the images are clear enough and not too noisy to declare that the information identifying McCain was not typed on the form at the same time as the other material on the form not specifically known to tie to McCain.
Beware eyeballing noisy data.