sounds of world english

Have you ever misunderstood someone or been misunderstood by someone who speaks with a different accent? The sounds that an American hears as 'Bob the clerk' may be heard by an Australian as 'barb the clock'.
 
Quantifying accents by analysing sounds is slow, expensive and limited. Here, we analyse how you perceive computer-generated sounds. This will show you the 'map' of your accent, in increasing detail as you listen to more sounds. By combining results, we are making 'maps' for different regions. To contribute to this global study, click begin the survey.

Overall, these two graphs are similar: Americans and Australians can usually understand each other! Look at some differences, such as "hood" and "haired".
 
For the Australians in this sample, the words "hud" and "hard" have a similar sound, the main difference is the length. For this sample of Americans, it is "hud" and "heard" that are distinguished by length. For an Australian, a long bud is a bard, for an American, it's a bird. (The current US sample is biassed by the number of Californians -- volunteers from other states can help redress the balance).
 
Within each group -- and even for one speaker -- there is considerable variation. The next graph shows the statistics for the first group we studied in detail.
 
 

This map shows the data for Australians, as at 26/09/2007. What does this graph mean?

The two most important parameters in determining different vowel sounds are the first two formants, which are frequency bands with increased power. These are the two axes on the graph. The axes are traditionally plotted backwards, as here, so that they approximately correspond to the axes long used by phoneticians and linguists: F1 (vertical) approximately corresponds to the jaw height (which correlates negatively with the extent of the mouth opening). F2 (horizontal) approximately corresponds to the position (forward or back) of the constriction of the vocal tract where the tongue is close to the roof of the mouth. Other important parameters are the length of the vowel and other formants. More about the science of speech.

"Short" on a vowel means that more than 75% of our subjects' selections were short and analogously for "long". This graph shows the map of how this group of subjects (a small number, early in the life of the study) recognise vowels with particular values of F1, F2 and vowel length. This graph probably corresponds closely to the graph for their accents, or for the accent of English they are most used to hearing. When we obtain enough (serious) surveys, we shall have maps for different countries and different regions, so thank you very much for participating!