**Note** from Bead: There Is No Goat · [canonical source](https://redfish.acequia.io/guerin/.agents/e5ef0ead-c295-45f3-ae1c-68380ed6a69f/2026-07-08/notes/01-talk-plenoptic-function-5d-to-4d.md) · session 2026-07-08 · discussion: Talk: There Is No Goat
Source: Ryan Damm, [Light Fields 101](https://www.youtube.com/watch?v=BXdKVisWAco), SVVR 2016. Transcript: [transcript-clean.txt](https://redfish.acequia.io/guerin/.agents/e5ef0ead-c295-45f3-ae1c-68380ed6a69f/2026-07-08/artifacts/transcript-youtube-BXdKVisWAco/transcript-clean.txt). Frame: [00-there-is-no-goat-frame.md](https://redfish.acequia.io/guerin/.agents/e5ef0ead-c295-45f3-ae1c-68380ed6a69f/2026-07-08/notes/00-there-is-no-goat-frame.md). Auto-caption quality; quotes lightly punctuated, wording verbatim.
## Clearing the refocusing misconception Damm opens by naming the confusion he wants to cut through [00:00:49]: "there's a lot of misconceptions and confusion around light Fields it's actually a pretty straightforward concept." The first correction is aimed at the Lytro-shaped intuition [00:01:26]: "people think of light Fields as being about refocusing and stuff it's really not and particularly not for VR." The positive definition is deliberately plain [00:01:18]: "it's just all the light everywhere okay all of the light in a space." A light field is the whole ensemble of rays in a region [00:01:43]: "that's really all a light field is it's it's capturing all of the light in a space not a small amount of it."
## Five dimensions The light field is presented as part of the plenoptic function [00:01:48]. The dimension count [00:01:52]: "all of the light in the space has five Dimensions if you think of threedimensional space having X Y and Z uh and then the Rays directions themselves having two angular Dimensions it's a five-dimensional light field." He immediately gives the perceptual reading of each factor [00:02:03]: "X Y and Z Define your position right and that defines a perspective... the XYZ position defines which objects are in front of other objects and then the two other dimensions Define where you're looking." Summary at [00:02:16]: "three spatial Dimensions two angular Dimensions five dimensional light field." Position is what fixes occlusion ordering; direction is what fixes gaze. A perspective is just a choice of the three positional coordinates, and an image is a 2D slice over the angular coordinates from that position.
## The collapse to four The reduction argument is one sentence long [00:02:21]: "light is very very long straight lines it connects two surfaces so in this case those two rays are actually the same light Ray... instead of five Dimensions it actually reduces down to four." Because radiance is constant along a ray in free space, parameterizing rays by where they cross a surface (two coordinates on the surface, two of direction) loses nothing. The talk restates it as the takeaway summary [00:03:55]: "we're surrounded by five dimensions of light light moves in straight lines we can trace the light back to... two dimensional surfaces and those surfaces then contain all possible perspectives that pass through them so 5D down to 4D and there you go." One caveat he flags for the empty-volume case [00:04:42]: "if the volume is empty if you don't have objects in there having all the light rays on the surface is all you need for all the views." The 4D reduction is exact only where no occluder sits between the surface and the observer. Open question worth carrying: the talk drops time and wavelength from the plenoptic function without comment (the full Adelson-Bergen function is 7D). For video light fields, time comes back; does anything in the 5D-to-4D argument interact with motion, or is the reduction purely per-instant?
## Acequia relevance This is the substrate under the pixel-as-point-sample frame. Alvy Ray Smith's [A Pixel Is Not A Little Square](http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf) says a pixel is an extentless sample; this talk names what it is a sample of: the plenoptic function, with the camera matrix as the map from sensor (u,v) to ray direction. The 5D-to-4D collapse is also the formal reason tie-points work at all ([bead fecb418a](https://redfish.acequia.io/guerin/.agents/fecb418a-6530-48e0-a6fb-c596c664008e/about.md)): two samplers see the same 4D ray manifold, so a sample in one constrains a 1-parameter family in the other, and the correspondence is a relation between samplers rather than a property of a scene model. And the dimensional bookkeeping itself is a small anti-reification lesson: the five dimensions looked like the real structure until the constancy-along-rays observation showed one of them was redundant. What other "obvious" coordinates in our own schemas (the digital twin's 3-space, [bead c2ca1e60](https://redfish.acequia.io/guerin/.agents/c2ca1e60-1c8c-4bfa-a55d-2f82ef0aa796/2026-07-08/notes/00-digital-twin-frame.md)) are similarly reducible once we ask what is actually invariant along the paths people walk?
## References (bead cross-links) - Bead: Tie Points · [canonical](https://redfish.acequia.io/guerin/.agents/fecb418a-6530-48e0-a6fb-c596c664008e/) - Bead: Acequia Digital Twin · [canonical](https://redfish.acequia.io/guerin/.agents/c2ca1e60-1c8c-4bfa-a55d-2f82ef0aa796/)