nPVI

nPVI, short for normalized pairwise variability index, is a statistic used to quantify rhythmic timing in sequences of temporal intervals, especially in speech prosody. It assesses how variable successive rhythm units—typically syllables or vocalic intervals—are in duration, enabling cross-language comparisons of rhythm.

Calculation proceeds as follows. For an utterance, obtain a sequence of durations d1, d2, ..., dm for consecutive rhythm units. For each adjacent pair (di, di+1) compute the absolute difference |di − di+1| divided by the mean of the two durations, ((di + di+1)/2). Average these values across i = 1 to m−1, then multiply by 100. The formula is nPVI = (100/(m−1)) × sum_{i=1}^{m−1} [|di − di+1| / ((di + di+1)/2)]. The result is unitless and typically expressed as a percentage. Higher values indicate greater relative variability between adjacent intervals; lower values indicate more even timing.

Origin and uses. The measure was introduced by Grabe and Low (2002) to study rhythm across languages and has since been employed to compare languages and dialects, often in terms of rhythm classes such as stress-timed versus syllable-timed. It can be calculated on different unit types (syllables, vowel intervals, or syllable nuclei) depending on the data and research question. While it is frequently cited in discussions of linguistic rhythm, the interpretation of rhythm typology based on nPVI remains debated and can depend on methodology and data selection.

Limitations. nPVI is sensitive to how rhythm units are defined and measured, the length of the data sample, and transcription practices. Comparisons across studies require careful control of these factors, and nPVI is best used as a relative index within a consistent dataset rather than an absolute measure of rhythm.