I am a grad student who wants to compare the prosody of two of my classmates from different regions who speak different dialects of English.
The obvious tool to use is ToBI, of course because it annotates pitch accent and boundary tone. But it looks like it requires a steep learning curve.
What are the key advantages to using ToBI over just "standard" f0 analysis in Praat? I wonder if the calculation of pitch is different in Praat vs ToBI, for example.
Thanks for the advice, y'all!
You can't use the standard f0 analysis. There's no such thing as standard. You need a framework e.g. Autosegmental-meterical Phonology (AM)
supposing you are going to work in the phonological category
But again this is when you are going to do a scientific study for a non-scientific observation you can take guess by looking at the pitch track but it's not scientific.
About prosodic transcription systems:
Actually ToBI is not the only system for transcribing prosody. There numerous systems within different frameworks available: INTSINT (Hirst, 1998), IViE (Grabe, 1998), Tilt (Paul Taylor, 1998), RFC (Paul Taylor, 1994)...
ToBI is one of them
The advantages of ToBI over other systems like INTSINT are as follow:
ToBI is easy to learn
One can easily learn ToBI you don't need special guides. only two letters H and L which if you happened to be familiar with AM Phonology you have seen them in other systems (in INTSINT). Some special characters like % and * need to be interpreted before using and that's easy supposing you have an understanding of AM phonology.
ToBI is easy to use
You can use ToBI right away, no need for special calcuations, no need for usage of other formulae. It's straight.
ToBI is easy to read
Even if your first time looking at a ToBI transcription you can easily guess what are the transcriptions and you can read them easily.
The disadvantages of the ToBI system:
Well, you can't always say that the easiest method is the best one. We still don't know what parts of prosody do we need so that we can throw away the others. ToBI is a system which neglects tremendous amount of prosodic data and that's not right.
In the end it's up to you to choose whatever works for you. But when talking about prosodic data, you have to be careful.
It’s true that there is a learning curve, but I’m not sure what the alternative is. What does ‘f0 analysis’ mean here? How can you interpret changes in pitch, intensity, duration without an interpretive framework?
The point of ToBI is to assign meaning to the physical observations (e.g., pitch, duration), which otherwise do not have meaning. To do this, pitch contours are interpreted within an abstract framework that says ‘this pattern means this’. I say that it is an abstract framework because it reduces the complexity of the signal by removing variability that is not meaningful in order to more clearly focus on the variability that is meaningful and contrastive for the language variety in question.
For example, we notice the final fall/final rise pitch distinction between statements and questions in English. For this contrast, (and speaking a little simplistically) it doesn’t really matter what the absolute f0 is, how loud the sound is, or what’s the duration of the contour; the relative change leads us to categorize the boundary as Low or High for this contrast. In order to have any confidence in our conclusions, we have to have a broad understanding of what prosodic patterns are typical and possible for the specific language variety, how actual speakers interpret those patterns, and a model that takes into account the interactions of all these patterns. This is the same as e.g. segmental phonology (‘why have phonemes when I have a spectrogram?’) or syntax (‘why have a grammar when I have words?’).
