What do humans learn in a double, temporal bisection task: absolute or relative stimulus durations?

The relative-coding hypothesis of temporal discrimination asserts that humans learn to respond to the relative duration of stimuli ("short" and "long"). The most frequently used procedure to test the hypothesis is the double bisection task. In one task, participants learn that red and green are the correct comparisons following 2s (short) and 5s (long) samples respectively. In another task, participants learn that triangle and circle are the correct comparisons following 3.5s (short) and 6.5s (long) samples, respectively. Later the samples of one task are tested with the comparisons of the other task, and vice versa. According to the hypothesis, participants will choose red following a 3.5s sample because that sample is short and red is the comparison that goes with short. Similarly, they will choose circle following 5s samples because that sample is long and circle goes with long. We replicated this procedure and improved it by introducing several sample durations during testing to obtain the whole psychometric function of each task. Results from Experiment 1 only partially corroborated the relative-coding hypothesis. Results from Experiment 2 did not corroborate the hypothesis. The combined data from Experiments 1 and 2 partially corroborate the hypothesis. Alternatively, we present an explanation of relative-coding-like results that posits exclusively absolute coding of temporal stimuli.