ASA/CAA '05 Meeting, Vancouver, BC

Visualizing the Canadian English Vowels

Robert Hagiwara - robh@cc.umanitoba.ca
Linguistics Department, University of Manitoba
Winnipeg, MB, Canada R3T 5V5

Popular version of paper 2pSC23
Presented Tuesday afternoon, May 17, 2005
Joint ASA/CAA Meeting, Vancouver, BC

Every language has dialects — forms of the language characteristic of particular social groups. Some of the most important and distinctive aspects of English dialects (or 'accents') have to do with the pronunciation of vowel sounds. For instance, in some dialects of English 'caught' and 'cot' are pronounced with different vowel sounds, and in others they sound the same. Accent becomes an important marker of national or regional identity.

Canadian vowels have previously been described mostly by careful listening. However it can be very difficult, even for trained experts, to simply hear very subtle differences in sound, and even harder to describe those subtleties in ways that can be meaningfully compared. The project I am presenting uses acoustic analysis to measure the physical properties the sound waves of speech in order to establish more objective and reproducible description of Canadian vowels. I hope in the future to use these data to compare what happens to these vowels in different contexts, rates of speech, and styles of speech. In this presentation, I am offering some preliminary findings, based on 10 younger adult speakers from Winnipeg, who were recorded reading a script of words of that illustrate the vowel sounds of Canadian English in a standardized context ('heat, heed, hit, hid' etc.).

Traditionally, we describe each vowel sound in terms of how we think the tongue is configured to make the sound — higher or lower in the mouth, more front or more back — and locate them impressionistically in those two dimensions. Many previous studies have established that these dimensions have acoustic correlates in the frequencies of the first several formants - those resonances that result from the shape and position of the organs of speech. I convert the acoustic measurements and represent the results in a way that mimics the traditional perceptual dimensions.

In Figure 1, the vowel symbols (in the International Phonetic Alphabet) represent the linguistic vowel category in its position in the 'vowel space'. You can see that the /i/ vowel (e.g. 'ee' in 'heed') and the /e/ vowel ('ai' in 'aid') are both very high and very front (represented toward the left of the figure) in the space. The /o/ vowel ('oa' in 'oat') is the furthest back in the space. The lowest vowels are /ae/ ('hat') and /a/ ('hot').

For the most part the results from my work confirm, rather than challenge, the traditional descriptions of Canadian vowels. For instance, in the lower right of the figure, the position of two vowels overlap almost completely — these are the vowels in 'cot' and 'caught', demonstrating that Canadian English (at least as spoken by these speakers in the laboratory setting) is in fact one of those dialects that do not distinguish these two sounds.

However, I have some findings which are more surprising. Canadians are well known to have different vowel sounds in 'lout' and 'loud', where in other dialects these sounds are usually the same. This difference, Canadian Raising, is also seen in the difference between 'ice' and 'eyes', and other pairs of words. Canadian Raising is thought to affect vowels that move from low-to-high in the vowel space. For the 'ou' in 'loud' the tongue starts out very low in the mouth, but 'moves' up and back (toward 'oo'). In 'lout', the vowel starts 'higher' in the mouth (closer to 'u' in 'but', which is the upside-down 'v' symbol) before starting its movement. This is how Canadian Raising is typically described: the starting point of these vowels 'raises' from 'a' to 'upside-down v'.

Figure 2 shows the current acoustic results for these sounds. The 'unraised' variants (/ai/ stands for the vowel in 'eyes', /au/ for 'loud') are represented as solid-lined arrows, representing the movement I have measured in the formants of the vowel sounds. The movement of the 'raised' variants (in 'ice', 'lout') are represented with dotted arrows. For reference, the moving vowels have been superimposed over the stable ones from the previous figure.

As you can see, the 'raised' sounds have starting points that are higher in the space than the unraised sounds. But that isn't all. From my data, it looks as though it's not just the beginning of the vowel that has raised, but the whole of the movement. And not only has the whole movement been raised, but it has also moved front or back, depending on the overall direction of movement. Thus I describe the difference not as 'raising of the starting point' but a change that affects the whole of the movement, in the direction of the movement. As far as I know, this has never been described. Another interesting finding is that this description also applies to the /oi/ sound (as in "void") in which the tongue moves from front-to-back, with the result that the whole of the movement is further forward in the space in the 'raising' context (e.g. in 'voice'). Thus I describe 'Canadian Raising' not as something specific to /ai/ and /au/, but general to this class of 'moving vowels', and not as 'raising' in the space but 'advancement along the path of the movement'. Again, I don't think this has ever been described, and I'm not sure it's something you can 'just hear' even with well-trained ears.

How general these findings are — whether the things I'm seeing are particular to this group of speakers, or general properties of Canadian English — must wait for further research. But I hope in my presentation to 'open the eyes (and ears)' of researchers working on Canadian and other varieties of English, and on the question of phonetic variation in general, to some of the techniques I am developing and the kinds of findings we might encounter.