Mechanisms for the coding of speech sounds may be suggested by a knowledge of factors not having cue value but affecting the discriminability of cues. Two experiments were run on adolescent listeners' identification of /bε, dε/ with near‐boundary values of syntheticF2andF3transitions; effectively, this measures discrimination. The trajectory ofF0over the duration of the formant transitions could be high falling or low rising but first‐pulse timing was controlled. Overall,F0had no effect; lowF0conditions gave intermediate performance, while highF0gave high performance for narrowF2bandwidths but low performance for broadF2bandwidths. These results are consistent with joint operation of peak‐picking and spectral‐shift modes of spectral analysis but not with periodicity analysis for theF2andF3region. Biasing effects indicating secondary cue value did emerge; for example both narrow bandwidths and highF0favored /bε/ responses. Such effects have to be explained partly as valid perceptual inferences to articulatory events and partly as byproducts of auditory transforms.