Phonetic Exercises

Phonetic Exercises - Languages in Danger

Author: Maciej Karpiński
Bon Ton(e)

Tones in languages

Ngwo is a Bantu language spoken in Cameroon. As many Bantu languages, it is a tone language, which means that the a change in the melody of a word may lead to a change in its lexical meaning or grammatical category. There are still many tone languages that have been not precisely described or only recently recognised as “tonal”.

Download the first recording on the list from this web page:

Listen to the signal but do not open it for viewing in Praat nor Wavesurfer.

After an introduction, you will hear a native speaker of Ngwo saying words in English followed by their equivalents in Ngwo (usually repeating them three times). Just listen (do not view any visualisations of the sound!) to the first five words (belly, shoes, dance, case, house).

Try to sketch how their melody goes (horizontal line for a flat melody section, a rising fo a rising melody, etc.) for each of these words. Just draw on a piece of paper. If you want to be more precise, you can use score paper but this is not necessary. Notice that all the repetitions of each word are very similar in respect to their melody which may mean that the melody itself is quite important here and cannot be arbitrarily changed.

Now you can open the signal in Praat (View & Edit window) or WaveSurfer (Speech Analysis mode).

In Praat, select [Show Pitch] from [Pitch] menu in the View & Edit window (if it is not already activated). Compare the drawings of the melody lines with what is now represented by the software that you use. Where can you find most differences?

[It is relatively easy to spot falls and rises although as non-native speakers we may face problems if we are to judge whether a fall or a rise is “meaningful” and it is not just a side-effect of articulation, not its consciously achieved goal. If there are, for example, high and low rising tones in a given language, you also have to analyse each of them in the context: How and when are they produced by a particular speaker? How large the variation can be?]

Try to remember the five words in Ngwa and record yourself saying them aloud.

Compare your pitch contours to the original ones. Listen to the original and your realisations in pairs. You may want to correct yourself and re-record your word list. Do not hesitate to do this!

[If your native language is not a tone language, it may take years to master fine reproduction of tones in your utterances. But here, with a limited set of five words, produced in isolation, it should not be difficult to achieve good results.]

Finally, open a corresponding graphics (jpg) file from the web page you downloaded the recording from.

The transcriber annotated tones by ear and marked them with diacritics above the letters in a very intuitive way. Compare his results with your initial drawings. Can you follow this way of melody transcription? If yes, take another five words from the original recording, try to transcribe them using letters that represent similar sounds in your language and, finally, mark the tones above the letters. Compare your results with the original transcription in the graphics file.

[For less known languages with the tonal systems that are not fully described, there can be significant differences between annotators and one can hardly say that one is wrong and another one is right. Once the phonological system is known and the inventory of possible tones is described, it is much easier. For example, you may wonder whether a given tone is a “low” or “extra low” but once you know that in a given language there is no such a distinction (there is only one “low tone”), you just mark it as “low” and accept the difference as resulting from individual variation that can be attributed, e.g., to an emotional state.]