This paper investigates a diachronic change that has taken place in the production and perception of the “rising” tone in Bangkok Thai, comparing data collected in the 1960’s and 1970’s with data collected 2004 to 2006. Regarding production, several studies have found that the abrupt pitch rise that had characterized the rising tone in earlier studies is now curtailed, such that the pitch of the rising tone does not actually rise above the middle of the pitch range. Concomitant with the change in the production has been a change in perception. Abramson (1978) found that a steeply rising trajectory, from the bottom of the pitch range at syllable onset to the top of the pitch range at syllable offset, was identified as a lexical item with a rising tone in 90% of presentations. In a partial replication and extension of Abramson’s experiment by Zsiga & Nitisaroj (2007), the same steeply rising trajectory was identified with the lexical item bearing the rising tone in only 5% of presentations; 85% of the time, this trajectory was identified with the high tone item. Zsiga & Nitisaroj found that the only tokens reliably identified as “rising” were those in which the pitch contour reached the bottom of the range at or near syllable midpoint, and that any actual rise is optional in citation form and prohibited in connected speech. It is argued here that this change can be understood only in the context of an abstract phonological representation, specifically, an autosegmental representation in which contour tones are compositional, and the mora is the tone-bearing unit in Thai. In this representation the rising tone has two pitch targets: L associated to the first mora, H associated to the second. The diachronic change can then be modeled as phonetic reduction of the H tone associated to the second (weaker) mora, followed by a shift of attention to the L tone associated to the first mora. Viewed through the lens of an abstract phonological representation, the diachronic change can be understood as a shift in perceptual importance from one phonologically specified target to another. Without reference to such a representation, the reversal in perception of rising trajectories makes no sense. This research thus supports the importance of abstract phonological representations in constraining the targets and outcomes of diachronic change in both perception and production.