Language and Gender

In addition to the simulations in the area of knowledge representation and processing I have done some experimental work in the area of language memory and processing. One result is a study on the question if speakers gender is discernible in transcribed speech.

Building a controlled corpus of dialogues

To investigate this question, a controlled collection of dialogues has been build. To do so, students performed an experiment in which one person had to explain another person how to construct a small model from parts of a construction kit without seeing each other. The gender distribution on the two roles (explaining and building) had been distributed equally. The dialogues have been recorded and transcribed.

Differences in Language and Perception

To find out if subjects are able to recognize the gender of speakers from the transcripts, different subjects received a sample of four sections of dialogues, one for each possible distribution of gender to roles, and were asked to indicate the gender of the speakers.

The results show a weak but significant tendency towards correct identification of speakers gender. But they show as well a significant stereotype: subjects tend to assume that the person who is explaining is male and the person who gets the explanation and is building the model is female. This is true although the subjects had been told that the gender distribution is equal on their samples. This stereotype was found only for female raters. I have published a more detailed report in "Sex Roles": Is Speakers Gender Discernible in Transcribed Speech? (Ferber 1995 [->]).