ISCA Archive Interspeech 2023
ISCA Archive Interspeech 2023

Nonbinary American English speakers encode gender in vowel acoustics

Maxwell Hope, Charlotte Ward, Jason Lilley

Encoding of gender in speech is a well-researched phenomenon, especially as it concerns men and women. Men tend to produce certain acoustic characteristics in certain ways (low fundamental frequency (F0), lower formant frequencies, lower center of gravity for [s] in English) compared to women, though these characteristics also differ based on other group memberships (e.g. race, sexuality, etc). Those who are more feminine, regardless of categorical gender, have been found to produce an increase in F0 and a larger vowel space. However, these previous studies used largely cisgender women and men or only examined encoding of binary gender in speech and did not consider encoding of "other" or nonbinary gender in speech. This study recruited American English nonbinary speakers to record 400 utterances and correlated acoustic characteristics with multidimensional gender. Masculine, feminine, and "other" gender are significantly correlated with vowel acoustics.