Step 1: Picture the instrument in use.
Think of the parallax bar as a measuring tool that sits on top of a stereo pair while you look through a stereoscope. It has two small etched marks, one on a fixed glass plate and one on a movable glass plate connected to a micrometer screw with a numbered drum. The pair of photographs is first taped down or clipped onto a base board with their flight lines collinear, and the stereoscope is placed over them so the observer sees a fused 3D model of the terrain.
Step 2: Recall why parallax exists between the two photos.
Because the camera station shifts by the air base $B$ between the two exposures, every ground point is imaged at a slightly different position in the two photographs, and this shift is larger for higher points and smaller for lower points. This along-track shift is called stereoscopic parallax, $p$, and is the raw measurement from which relief (height) is derived using $h = H - \dfrac{fB}{p}$, where $f$ is the camera focal length and $H$ is the flying height above the datum.
Step 3: Connect the instrument reading to the measured quantity.
When the observer adjusts the micrometer screw so the floating mark formed by fusing the two etched marks appears to rest exactly on the ground point in the 3D stereo model, the bar's scale reading is the parallax of that point. Taking such readings at two points, say the base and the top of a tall object, and differencing them gives the differential parallax $\Delta p$, which feeds directly into a height computation of the form $\Delta h = \dfrac{H\,\Delta p}{p_a + \Delta p}$ for the relative height between the two points. This is the whole reason surveyors carry a parallax bar into the stereo-plotting room.
Step 4: Rule out the distractors.
Flying height comes from GPS/altimeter data recorded during the flight or from known ground control, not from the bar. Photo scale is $S = f/H$, a ratio computed separately from focal length and flying height, and the focal length is a fixed camera constant printed on the calibration report; neither of these quantities is something a parallax bar physically reads off the stereo model. The bar is purpose built to read one quantity only, the stereoscopic parallax of a point.
\[ \boxed{\text{parallax}} \]
