The multilingual vowel perception research project is a study of the reactions of speakers of different languages to the same set of synthetic vowels. The purpose is to see how different vowel systems are reflected in the answers to this set, which includes strange vowels for many languages. Of special interest is the question of possible "universal" vowels, vowels which could be called the same in several languages. In time the study will yield information about individual and dialectal differences, effects of language contacts and it is also interesting from the point of view of the testing technique.

The test consists of 386 synthetic vowels which cover the entire vowel space except for diphthongs and nasal vowels. To produce the vowels the Klatt synthesis program was used and the results checked against the Praat analysis program. The vowel space has been created by varying F1 from 250 to 800 Hz with steps of 30 mels and F2 from 600 to 2800 Hz with steps of 40 mels. F3 is 2500 Hz as long as F2 is 2000 Hz or below and higher by 200 mels when F2 is above 2000 Hz. There are two different sets according to the duration of the vowels, long vowels of 350 ms and short vowels of 120 ms. The pitches in the two sets can be seen in the scheme below.

	Long vowels			Short vowels
Time (ms)	0	120	350	0	40	120
Frequency (Hz)	100	120	80	100	110	90

In the test the subjects are asked to identify the test vowels as vowels of their native languages. They are also asked to rate the goodness of the vowels on a scale from 1 to 7. A vowel can also be repeated any number of times and the answering time is up to the subject. Most subjects have taken about 40 - 50 minutes for one test.

The first tests were made in the computer class of the Language Centre of the University of Turku, but now the test is also available in the Internet (http://www.utu.fi/hum/fonetiikka/vowel.html). There have been some difficulties with the system, especially with Macs, but we have also received complete test answers from both Finland and abroad. We have results from about 200 subjects representing about 20 languages, Finnish informants forming the largest group, of course. We hope to get many more answers to the test to make it serve the multiple goals set for it.