We will next consider F3, which is the 5/4-wavelength standing wave in the idealized model. F3 is known to be involved most saliently in the contrast between English /r/ and /l/ (in some dialects). Retroflex and lateral continuants may be distinguished on spectrograms2.9in this way: [] has a severely lowered F3, while [l] is often distinguished from it by a raised F3. While this is not true of all sounds labelled /r/ or /l/, this relationship is nonetheless a striking mystery of acoustic phonetics. How is F3 lowered so severely in rhotic sounds? We may predict from the acoustic fact that there is a constriction at one or more antinodes of the 3rd resonance. The prediction is resoundingly true, as discussed in detail in Ohala (1985).2.10 It turns out that [] has a constriction made either by the tip of the tongue curled far back, or by bunching and raising the tongue, so as apparently to approach the middle antinode, A6, of the 3rd resonance. Additionally, retroflex continuant /r/ in English occurs with a seemingly peculiar, localized constriction in the pharynx: ``There is also a constriction in the pharynx below the epiglottis''(Lindau, 1975:27). If this occurs at the F3 antinode, A5, then the effect would be the observed one, a lowering of F3. Finally, it is well known that retroflex /r/ in English has lip-rounding. Thus not only is there a constriction at one antinode, there are constrictions at all three of the antinodes of the F3 standing wave. This is striking confirmation of the theory of the relations between acoustics and articulation.
Why should F3 be raised with a clear [l]? The theory predicts a constriction at nodes N5 and N6; contact of the tongue-tip behind the teeth during [l] would seem to support this prediction. Given the number of precise and accurate predictions made by Rayleigh's node-antinode rule, we may have guessed that those [l]'s with raised F3 also have a localized dorsal constriction, at the location of node N6, as it appears they do.
The concomitant rounding of the lips which accompanies back vowels is quite similar to the simultaneous multiple articulations of [] and [l]. To lower F2, not only can there be a constriction at A3, there can also be a constriction at A2, namely at the lips. It is no coincidence therefore that back vowels, which have a lowered F2, should in most cases also have lip-rounding. Both articulations amount to antinode constrictions, and they have the same effect on F2. With both backing and rhoticity, completely unrelated articulations cooccur; the constrictions happen to be those which have the identical acoustic effect. This suggests that the acoustic effect is primary, and the articulation simply follows from the desired effect. A constriction at both antinodes results in a lower F2. The same holds for F3. Even when the constrictions are far apart, their effect is the same, and if the articulation simply follows from the desired acoustic effect, the separate constrictions that have no apparent articulatory connection make sense. Without the acoustics to make sense of the articulations, the cooccurence of these independent constrictions is a mystery.