Originally Posted by Matthew 
Macquarie Dictionary: "depth of field, n. Optics. The range of distances along the axis of a camera or other optical instrument, in which an object will produce a reasonably clear image."
[...]
After all, the lens is still projecting exactly the same image on exactly the same area of sensor
With this bit of preparation, the core idea of the definition of DoF can easily be worked out (and I don't agree that the term's meaning has changed over the past 100 years -- it is carved in stone and teached in optics classes; the interweb sometimes has problems to keep simple things simple
).
1.
Lens projecting exactly the same image: Yes, and out of context, DoF would always be
zero. Optically, only
one subject plane is in focus and all other parts of a subject are not sharp.
2. Therefore,
reasonably sharp is the keyword here.
3. As this term needed a precise definition, an otherwise sharp point is allowed to smear out over a finite-sized area, the
Circle of Confusion (CoC). And its size matches the human eye's capabilities. Typically, the Zeiss formula is used:
CoC = image diagonal / 1730
So, whenever the image diagonal changes (e.g., when cropping), so does DoF. Note that pixelpeeping is cropping because your monitor is limited in size.
If you go thru all the math, you'll see that independently from sensor size, focal length and everything else, DoF
only depends on three factors:
- The distance to the subject [m]
- The physical diameter of the lens' aperture in millimeter [mm]
- The Field of View (FoV) of the image, in degrees [°]
Of course, the
2 parameters, FoV and lens' aperture in millimeter can be computed from
3 variables (sensor size, focal length and f-stop). However, this obscures things and sometimes leads to endless debates