The effects of prosodic factors on the spectral rate of change of vowel transitions are investigated. Thirty two-syllable English words are placed in carrier phrases and read by a single speaker. Liquid-vowel, diphthong, and vowel-liquid transitions are extracted from different prosodic contexts, corresponding to different levels of stress, pitch accent, word position, and speaking style, following a balanced experimental design. The spectral rate of change in these transitions is measured by fitting linear regression lines to the first three formants and computing the root-mean-square of the slopes. Analysis shows that the spectral rate of change increases with linguistic prominence, i.e., in stressed syllables, in accented words, in sentence-medial words, and in hyperarticulated speech. The results are consistent with a contextual view of vowel reduction, where the extent of reduction depends both on the spectral rate of change and on vowel duration. A numerical model of spectral rate of change is proposed, which can be integrated in a system for concatenative speech synthesis, as discussed in Paper II [J. Wouters and M. Macon, J. Acoust. Soc. Am. 111, 428-438 (2002)].