Lawrence D. Shriberg and Gregory L. Lof

Shriberg and Lof do what according to their extensive lit review is only the second largescale study of reliability data--largescale meaning not only "involving many participants or elicitations", but "looking at both intra-and inter-rater reliability, under many conditions, with different types of speech, with familiar and unfamiliar speakers", and etc. for about five pages of variables. They find that broad transcriptions are within acceptable levels (defined for no reason that they get into except tradition as at least 85% reliability) and narrow transcription is not, which as well as a neat backhanded reminder for phonology students like me of the importance of suppressing our own assumptions--look what happens when your own phonological system is no longer doing the work for you, i.e. when you have to transcribe the narrow unpredictables of e.g. "beet" instead of just /bit/--but also a more problematic conclusion for people doing clinical transcription and diagnosis, who then have to grapple with how to trust their data and its implications for clients' lives instead of just getting trained transcribers with deep familiarity in the language (not that this is possible in every language even for those doing frontline linguistics). they also, fascinatingly, find a significant effect for familiarity--if you know the person being transcribed, it pushes you this way or that--whether because they are clinical and you're rooting for them, or because you've heard them say something before and you're like "oh! I know! this is like that time!" and another transcriber is not, or I dunno why.


My big problem with this paper is the way it privileges acoustics--I get that you need a steady standard of comparison, but self-evidently, what ultimately matters in language as opposed to speech science is how you're heard and whether we can normalize your oddities--not what the instruments say you're doing. but this is obviously still a problem and call for training and data monitoring, and my quibble is just a rhetorical or manner thing.


Which is always where I gravitate, man. Such a humanities jerk at heart. This year off will be a good chance to evaluate my move into the sciences further. I mean, I'm doing fine, but the kind of rigour that they approach stuff like this with, the endless reams of data when I'm all "get to the discussion! let's go sublime!" makes me wonder about broader temperamental compatibilities. Ah well. One master's degree at a time. Clinical Linguistics & Phonetics.… (más)