A dataset containing 14,424 formant measurements from 1,049 vowel tokens. These came from generated nonce words. Formants are extracted at 21 points along the duration of the vowel. This dataset is relatively clean and can be used for vowel formant example data.



A data frame with 14,424 rows and 13 variables:


a unique identifier for each vowel token


the start time for that vowel


the end time for that vowel


the time where formants were extracted


how far into the vowel's duration (in terms of percent of the duration) the formants were extracted. 0 = onset, 50 = midpoint, 100 = offset


the F1 measurement


the F2 measurement


the F3 measurement


the F4 measurement


the generated nonce word I read


the consonant(s) before the vowel (if any)


the vowel class, in Wells' Lexical Sets


the consonant(s) after the vowel


I generated 333 nonce words of the form (C)CVC(C) with the following sounds:

  • Onsets: /t/, /d/, /s/, /z/, n/, /h/, / /, /st/, /sn/

  • Vowels: /i/, /ɪ/, /eɪ/, /ɛ/, /æ/, /a/, /ɔ/, /oʊ/, /ʊ/, /u/, /aɪ/, /aʊ/, /ɔɪ/

  • Codas: /d/, /z/, /dz/.

Generally, only coronals were used. Clusters were okay. Codas were only obstruents. All American English vowels were used. Not the most scientific dataset, but sufficient for my purposes.

Every combination of these levels was generated and repeated three times. The resulting list was sorted randomly. I read them in a quiet environment, manually aligned them, extracted formants using a Praat script (4 formants at 4500 Hz), and filtered out the bad measurements.

The result is a pretty clean dataset showing my vowel formant trajectories, in the environment of a coronal consonant.

The intended purpose of this data is so that I can quickly have a nice sample at my disposal when illustrating R functions. However, you may use this dataset however you please.

Metadata about me: White male, born in 1989 in suburban St. Louis where I lived until I was 18. Parents are from upstate New York and Minnesota. Lived in Utah, Brazil, and Georgia as an adult. Data was recorded July 2020 (age 31).