Part 1 2 3 4 5
During 2017, two photographs were highlighted in the news of Old West characters: the Sanpete Railroad photograph – proposed as a possible depiction of Butch Cassidy and the Sundance Kid – and the even more recent news of a new photograph believed to be Billy the Kid. Today, I will be briefly discussing these photographs and the facial alignment elimination stage of identifying what I call ‘sparse historic photos.’
Applying an alignment elimination stage
The act of assessing the identity of a person in a historic photograph is more difficult and less glamorous than those of us who attempt it might wish. As a scientist, one may have an intuition, but then one must substantiate it to a very high standard indeed. My authentication of the Dowe photo of Butch Cassidy and the Sundance Kid took several hundreds of hours of work over two years and I continue to find new ways to refine my methodologies. [1]
A very good first step to reduce the risk that this intensive research may end in failure is to do an ‘elimination’ step, in which the photograph either passes and is therefore worthy of further research, or does not pass, in which case no further investigation is required. (For anyone wishing to do this type of alignment analysis, I heartily recommend Joelle Steele’s “Face-to-Face” epublication [2] as a great place to start.)
Let’s consider this elimination step as it does or does not apply in three recent examples of sparse historic photographs: the ‘Sanpete Railroad’ photograph [3] considered to possibly be Butch Cassidy and the Sundance Kid, the ‘Billy the Kid croquet’ [4] and the ‘Billy the Kid Flea Market’ [5] photographs, and lastly the Dowe photograph of Butch Cassidy and the Sundance Kid I authenticated in 2016. [6]
Sanpete Railroad photograph
There was a stir in March this year about a photograph possibly including the Sundance Kid and Butch Cassidy working on a railroad. The photo was discovered by Utah social studies teacher Chris Vorhees while doing research in the Brigham Young University archives. Mr Vorhees came across the photograph in a large collection of photographs taken by George Edward Anderson, a travelling photographer who worked in the American West in the late 1800s / early 1900s. The character suspected of being Harry Longabaugh (The Sundance Kid) was presented with the known photographs using a form of horizontal alignment elimination. [3]
I’m sorry to say this, because I have no doubt that the person who did this analysis was well-meaning, but the depiction of the alignment elimination step published for the Sanpete Railroad photo is so rife with mistakes as to make it completely invalid as a whole. The following explanation is not meant to criticise, but to explain and hopefully to offer useful guidelines to anyone attempting these types of comparisons in future.

In the first instance, the preparation of the images for comparison is faulty in that there has been no attempt to align each face horizontally with the necessary precision, and it seems likely they have not been scaled to eye width match though this is a bit difficult to discern given the small facial sizes (which should fill the frames) in the images.
Lines 1, 2, and 7 are invalid as they do not use actual physical features as markers; hats interfere with lines 1 and 2, and line 7 has no physical reference at all. The remaining 4 lines – 3, 4, 5 and 6 – do not appear to align well [7] but it is difficult to judge due to the blurriness of the lines. It is possible to see a lack of alignment in lines 3 and 5 for the Sundance Ft Worth photo.
Also, no vertical alignment analysis was found, which is important, as it is part of the rigour of this elimination step; if this exists I would be glad to see it.
Suspected Billy the Kid photographs: at play at croquet and in a gang of Regulators
Two photographs suspected of being of Billy the Kid that have been in the news lately are a) the ‘BTK croquet’ photo owned by Randy Guijarro and his wife, having bought it in a Fresno memorabilia shop, and b) the ‘BTK flea market’ photo purchased in 2011 by North Carolina lawyer named Frank Abrams and hitting the news last month. I thought I’d try doing an elimination alignment on these photographs, comparing it to the only authenticated photograph so far (bought by William Koch for 2.3M in 2011). However, very quickly I found that it couldn’t be done.
I don’t have access to high resolution photographs of the unauthenticated croquet photograph. However, it is unlikely that even the highest resolution scan will be sufficient to even complete a credible elimination stage for the ‘croquet’ photo, let alone an authentication, since the print is only 4” x 5” and the area of the photograph containing the face of the proposed Billy the Kid figure is only 0.022% of that photograph. [8] You can easily see the problem in the figure below: once the face of the person in question is zoomed large enough to see in comparison with the authenticated face, little detail can be seen, and such as it is, it is very difficult to see any detail at all in the bottom half of the face, as though the person was moving his mouth while the photo was taken – perhaps talking?

From left: Known Billy the Kid photograph, ‘BTK croquet’ and ‘BTK flea market’ photos. Lack of resolution and image problems make it impossible to do an alignment elimination stage
The photograph announced this year is almost as problematic, though in a different way. In the new photograph it seems possible that the person moved their head sideways as the photograph was taken. This is suggested by the area between the eyes, the nose, and the philtrum below the nose, as well as the Adams apple, which all seem to have a slight shadow exposure causing the face to appear unnaturally wide. If this is the case, it means that any authentication would have to first attempt to reassemble the face, a process that would corrupt the pixel data of the image. I am unclear on the origin of the pink cheek colouring (I assume that this was painted in, perhaps by the photographer’s studio) but obviously this was not an enhancement to the appearance of the photograph, and probably obscured detail in the original image.
I would not venture to state that these proposed Billy the Kid images are or are not Billy the Kid because there is simply not enough image data (size, crispness and detail) to even complete an elimination stage assessment. Unless the ‘BTK croquet’ photo original contains more information than the versions commonly available online, which would in any case be complicated by the structural issues described above, I question if it can be authenticated. [9] The issues with the ‘BTK flea market’ photo seem highly confounding, but the original image / higher resolution scan may offer more scope for investigation.
Butch Cassidy Sundance photograph by O.S. Dowe
The alignment elimination stage images that I released in Brian’s Butch & Sundance: The New Evidence epublication were largely inspired by Steele’s groundwork in this area.
The intent of this first step was to determine if the suspected identities of the subjects in the Dowe photograph must be discredited immediately due to basic lack of facial similarity to the known photographs. This is the first of 4 steps I undertook in authenticating the identities of the men in the Dowe photograph. For each subject, the face portion of the Dowe and reference photos were placed together in a single comparison window. Each face was rotated as needed to bring it into horizontal alignment. To reveal any obvious differences in the composition of each face, each face image was scaled (size adjusted) so that the eye widths were as similar as possible between all images. The proportions of the face were then compared visually by overlaying guides at key points of the face. [10]
You can see the individual alignments in the video, however, today I thought it would be interesting to offer the two alignments (horizontal and vertical) placed one on top of the other. This also demonstrates how the two alignments cross-validate each other.


You can see how each image is slightly rotated – this is to (as much as possible) position the faces aligned horizontally with one another. Another thing to note is how the eyes are (again, as much as possible) all of similar size; the cropped images were carefully scaled to achieve this outcome so that the size and proportions of the faces relative to each other could be assessed.
Both subjects passed this alignment elimination stage.
Tomorrow
Tomorrow we will hear from my guest blogger, the man who discovered the Butch Cassidy and Sundance Kid cabinet card: my brother, Brian Mida Bleecker.
References
[1] Modern facial recognition that is so much in the news is based on the existence of many photographs of the same individual. Sparse historic photographs do not fit in this mold, and have to be approached with an intensive, ‘hands-on’ approach.
[2] Steele, Joelle. (2014) Face to Face: Analysis and Comparison of Facial Features to Authenticate Identities of People in Photographs http://www.facecomparisons.com/press-face-to-face.html Accessed 20 December 2017
[3] http://sanpetemessenger.com/2017/03/30/were-butch-and-sundance-among-sanpete-railroad-roustabouts/ Accessed 20 December 2017
[4] Little, Becky. (15 Oct. 2015) Does this photo really show Billy the Kid? It’s a $5 Million question. National Geographic https://news.nationalgeographic.com/2015/10/151015-billy-the-kid-photo-outlaw-photography-history/ Accessed 27 December 2017
[5] Fortin, Jacey (Nov. 16, 2017) A Photo of Billy the Kid Bought for $10 at a Flea Market May Be Worth Millions. New York Times. https://www.nytimes.com/2017/11/16/us/billy-the-kid-photo.html Accessed 20 December 2017 and ] http://www.history.com/news/history-lists/9-things-you-may-not-know-about-billy-the-kid Accessed 20 Dec 2017
[6] Bleecker, Brian (2016) Butch & Sundance: The New Evidence. mazon.com/dp/B01N3NQRRW
[7] Keep in mind that the original images are not prepared well, so alignment or lack thereof will likely be different with better prepared images.
[8] Full image released publicly is 1000 x 600 pixels and the area of pixels of the face in in question in this photo is 11×12 pixels approximately. This means that the face occupies (11×12) / (1000×600) pixels, or 132/600,000 pixels. This is equivalent to 0.022% of the area of the photograph.
[9] My consideration of these two photos was limited to the alignment elimination stage only. In considering them further, other factors must be considered alongside the forensics questions. For example, it is purported that close associates of BTK are depicted alongside the figure in question. Should these cases be proven it obviously lends support to the claims.
[10] Horizontal: pupil line, subnasale (point of bottom median attachment of the nose), stomion (labial fissure), supramentale (chin furrow), and pogonion (chin boss). Vertical: Exocanthions (outer eye corners), endocantions (inner eye corners) and median line. Note that in the case of Subject B (Butch Cassidy) the pronasale (tip of nose) is out of median align due to the aspect of his facial pose (looking to his right instead of facing forward like his comparator photos).
Copyright BMBC = Copyright Brian Mida Bleecker Collection