Abstract. This work investigates the selection of horizontal confined tracks for fission-track modelling. It is carried out on prism sections of Durango apatite containing induced tracks with mean lengths of ~16, ~14, ~12, and ~10 μm. Suitable tracks are identified during systematic scans in transmitted light. The explicit selection criteria are that the tracks are horizontal and measurable. We measure the length, width, orientation, and cone angle of each selected track and in some cases other dimensions. The confined track selection is in the first place dependent on a threshold width and in the second place on the requirement that the tracks are etched to their ends. In most cases the first condition implies the second, which decreases in importance as the tracks are shortened following annealing. The widest confined tracks, which must also be the shallowest, come to intersect the surface and are excluded. In general, the selection is dominated by the width of the etched tracks. This, in turn, depends on their orientation relative to the c-axis and the apatite etch rates, and their effective etch times. Despite the different geometrical configuration of the unetched host tracks and confined tracks, neither the angular distribution nor the etch time distribution of the confined track sample depends on the degree of annealing. This illustrates the general principle that those entities are selected that have the right properties for being selected. In this case etching-related factors determining the track width are the most important, while the known geometrical biases are second order. The track etch rate exhibits no demonstrable variation along the track, but signifi-cant differences from track to track. Moreover, although the track etch rate of induced tracks is not correlated with the extent of partial annealing, it is on average twice as high as the value for fossil tracks. Our length measurements are in good agreement with the annealing models for this apatite and etch protocol. We submit that this is not fortuitous and that it is possible to select a representative confined track sample, and perform reproducible and meaningful confined track length measurements. Deliberate or inadvertent biasing, carelessness or inexperience will of course give different results, but these should be treated as statistical outliers, not as an indication that track lengths are fluid.