The Trindade Island Photographs, 16 Jan 1958

Martin Shough

A study of the image-inversion hoax hypothesis
(Feb 2004. Last update 16 Mar 2004)

(index) Part One (Go to Part 2>)


Photo 1

Photo 2

Fig.1. Extreme (~100x negative scale) enlargements of the images from the first two of the sequence of four photographs, size and orientation adjusted to match. These are referred to as P1 and P2 below.
Original scans from first generation prints by A.J.Gevaerd, courtesy of Kentaro Mori, processed to enhance contrast and sharpen textural detail.)

the prima facie 'UFO' scenario

The 'naive' interpretation of these two images is that P2 appears to expose more of the underside of a roughly symmetrical lenticular greyish object encircled by a broad, dark flange. (Fig.2)


This would be qualitatively consistent with the perspective change implied by the photo sequence and by the testimony: The object approaches right to left across the sea and over the island (as in the sketch in Fig.3 below), doubling in elevation angle from roughly 8 to 15 degrees above the horizon. Considering (for simplicity) the flange as a disc maintaining itself level with the surface of the sea, its near edge would be elevated relative to the line of sight by an amount in the order of 10% of its own angular width due to the perspective change between P1 and P2. We can confirm by rough inspection that this is not inconsistent with Fig.2.

(Note: This doesn't mean, of course, that there are no other variables. There is testimonial evidence that the flight attitude of the disc was not consistently level, for example. There is also reason to expect measurable lateral motion-blurring at the order of speed reportedly estimated by the Brazilian Navy Intelligence Service analysis of 1958. For instance, a 100' object travelling at 700 mph would be displaced by 8% of its own length during a 1/125 sec. exposure.)


the prima facie inversion hypothesis

As originally suggested in 1958 by Capt. Sunderland, the US Naval Attache in Brazil, the image in P2 might simply be an inversion of P1. When rotated as in Fig.4 below these images are clearly somewhat similar. In particular, the region of the 'dark spot' on the top part of P1 and the similar but less dense region on the bottom part of P2 now appear in corresponding positions. These regions even appear to have a similar asymmetrical shape, best discernible in P2. It can also be argued that what previously appeared to be the underside of the dark flange, revealed in P2 due to the perspective change, now corresponds with the 'shadowed' underside of the object in P1.

Fig.4. Comparison of P1 with P2 rotated by 180

This is the basis of the hypothesis that the photographs are faked using different versions of a single image rotated 180 in the plane of the photograph. Leaving aside the mechanism by which this might have been done, the logic of the argument is simple: It is implausible that a 3-dimensional object would have globally rotated in the plane of the photograph so as to invert its axis of symmetry whilst projecting a locally self-identical image at the camera; but this is exactly what happens if you rotate a flat image in the plane of the photograph. The simplicity of the hoax hypothesis is made the more appealing by the fact that there is no explicit description in the witness testimony of the object inverting its axis of symmetry. Moreover, none of the four photographs happens to have captured the object with its axis of symmetry far from the vertical.

However although a basic structural similarity between P1 and P2 can certainly be claimed, the similarity is not an identity. It is evident to the eye that they are not simple inversions of the same image. Most obviously, the region of high density on the right side of P2 has no counterpart in P1, whose principal region of high density is on the left. The gross outline is also subtly different. Some similarities and dissimilarities are brought out by the edge-enhancement in Fig.5 below.




Fig.5 Edge tracing filter applied to P1 and P2, vertically stretched (x2) to exaggerate internal structure.

The hoax hypothesis is therefore not quite straightforward. If P1 and P2 do have an underlying identity, this is obscured by some other transformations in addition to simple inversion, and it then becomes difficult to separate hypothetical transformations that might have been applied to a photographic image from transformations that might have been undergone by a dynamical physical object of uncertain nature.

In short Fig.4 does not test between the hypotheses that are of interest to us, because it cannot tell us whether the transformed underlying identity belongs to a real object over Trindade or to an image in the darkroom. It does not amount to proof of the latter without supplementary assumptions that we now need to make explicit:

  1. that the apparent concealment of the dorsal 'dark spot' and the subsequent revealing of a similar ventral spot between P1 and P2 is only explainable as the migration of one permanent physical feature;

  2. that this implies a space rotation of the object, or of the image, in the plane of photograph;

  3. that such a space rotation of the object is physically implausible and/or testimonially unsupported.

(Unfortunately each of these points is arguable and takes us into realms of conjecture and interpretation. We return to these matters later. Additionally, of course, we would like to be able to show how the images might have been produced, and if the inversion is an accidental or a deliberate artefact of the process used to fake the pictures.)

For clarity, let us recall that one of the merits of this hypothesis, as it originally recommended itself to Capt. Sunderland in 1958, must be that it provides a simple explanation of a simple inconsistency - the hoaxer inserted one of his images upside down, by accident or to disguise a give-away identity. However as soon as one looks more than cursorily at the images, with the advantage of enhanced enlargements, it becomes clear that neither the apparent inconsistency nor the hypothesis invoked to explain it are as simple as they appeared. The images are not the same, even rotated. Also it quickly becomes apparent by experiment that no global processing of the P1 gamma will transform it into P2, or vice versa. The distribution of dark and light areas, whilst broadly similar, is in fact different.

'Well, this not so strange,' we may then temporise. 'The fact that the image is changed between P1 and P2 is only what one might expect if some complicated process has been used to fake a convincing sequence from a single source. The nature of the difference then probably tells us something about the method used.'

But this is open to the charge of being an ad hoc patching-up of an hypothesis which has been falsified, and at this point it is no longer quite clear whether our objection is that the images are too much the same, or that they are not the same enough.

closer examination of the inversion hypothesis

The status of the hoax hypothesis is not clarified on closer examination. Looking at the stretched images used for the edge enhancement in Fig.5 above we find the interesting internal linear structure shown in Fig.6 below.

(a) 2x vertical stretch, 180 degree rotation of P2 in image plane


(b) 2x vertical stretch, inversion of P2 normal to image plane (flip)

The vertical stretching elongates the fine grain structure identically on both P1 and P2, but this is quite distinct from the striated macrostructure which, in P2, can be seen to make an oblique angle with the vertical.

Fig.7 2x horizontal stretch of P1 and P2. Elongated grain structure is visible, but no lateral streaking.

In Fig.7 the two images are stretched by the same percentage horizontally for comparison, and although horizontal grain elongation is now obvious the oblique off-vertical structure of the P2 image identified in Fig.6 is not destroyed, but on the contrary remains very perceptible here. This demonstrates that the striation is not an artifact of the stretching process but is a real property of the scanned images.

Fig.8 2x vertical stretch. Difference in mean angle of striae, illustrated in rough outline tracing on the right. The anisotropy in P2 clearly runs across the direction of stretching indicated by the elongation of the film grain.

So the images carry this internal 'fingerprint' which can be compared, and we find that the result confirms the conclusion from Fig.5. Not only is the gross density distribution of P1 different from that of P2, there is a subtle linear anisotropy in the images which differs measurably. With these results in mind we can re-compress the vertical axis to the original ratio, as in Fig.9 below, and reassure ourselves that the features we have discovered in enhancements can be confirmed by inspection of the original images. The relative obliquity of the striae measured in Fig.8 is here restored to its proper value, which is evidently about 40.


This result can only remain consistent with the hypothesis that P2 is an inversion of P1 if the anisotropy is introduced differentially to both P2 and P1 after the inversion has been prepared. That is to say, the striation itself cannot be a by-product of a process of reproduction of P2 from P1 (or vice versa) since it exists in both images, both 'original' and 'copy'; and one cannot be produced from the other except by a skewing which would distort the gross outline of one image in relation to another (Fig.10) in way that we do not see.

Fig.10 Outline tracings of P1 and P2. On the left, as found; on the right, with the skewing necessary to allign the internal structure. Although neither P1 nor P2 is itself perfectly symmetrical, it is not possible to transform one into the other.

Go to Part 2>

All image processing by Presto!ImageFolioLE