PTLC 2017, part 2

Find below a brief report on the two keynote presentations on day 2 at UCL's PTLC conference. As usual, you will see info from the presentations interspersed with my own comments. Any errors in interpretation are my own!

Jim Flege - The cross-language acquisition of stops differing in VOT: Historical review and key findings 

It was truly exciting to see THE Jim Flege in the flesh, and I am not going to include much factual information on the presentation, as it is freely available here. I think I would like to mention a few things I have found really useful and/or comment-worthy in Flege's very interesting presentation on VOT acquisition and learning in different languages.

First, it came as a bit of shock to hear from Flege that his interest lies with pronunciation acquisition in "naturalistic" contexts, that is, his research is not necessarily concerned with instructional settings (where, funnily enough, many of us have studied/applied his models), but with how the Speech Learning Model works in contexts where the L2 is spoken (and yes, this is a common phenomenon, much of the research we cite is for second language in immersion contexts, rather than foreign language pronunciation learning in places where the L2 is not spoken outside the classroom...).
Flege makes an interesting criticism to experimental work:

It was really inspiring to see Flege criticise his own past research, and to see what "mistakes" (sic) he has made in past papers (e.g. 1987) regarding data collection, participant selection, and conclusions drawn. I think that only a true respected academic can really admit that things may not have been done correctly (unknowingly at the time), and the ability to look back and reflect using the current state of knowledge is really brave, and also promotes good, honest science. Part of this review of past research included the current need to displace the role of age of acquisition and the Critical Period Hypothesis to foreground the role of L1 input.

Flege highlights the role and importance of L1 input ("good input from native speakers", I quote) in the acquisition of VOT contrasts and timing. I totally see the point of native speaker input, and I support it, but I wonder if this view somehow undermines all our own input as non-native speakers (I did not ask this question, so I cannot really say if this is what Dr. Flege actually implied...again Flege's concern is with language learning "in situ", as it were). I think that even though our daily input in lessons may not be native-like, we also provide instructional input, based on our own knowledge and experience of what it is to have learned the phonology and phonetics of a language other than our own, so...complex stuff.
Flege proposes that there should be a 10-year period  (!) of native speaker input for phonological features to be fully acquired and formed (in particular with relation to VOT), since the formation of categories in L1 is also a lengthy process.

Flege also reminds us that the acquisition and formation of phonetic categories does not need then to be related to age, nor to a "special ability"

I also found interesting that studies on the "loss" of L1 features have been carried out, and yes, I believe I may have lost the naturalness of some of my own Spanish features (I would say in terms of intonation, especially).

Please refer to the slides in Flege's website for a review of all the research on VOT presented, which I obviously cannot do any justice to!

Dominic Watt - Private Ear training: Phonetics teaching for the next generation of forensic speech scientists  

Watt discussed something I have found fascinating for years: forensic speech science. He discussed some principles and features of this way of doing phonetics, such as the fact that a great part of what is done includes applied sociophonetics (synchronic and diachronic variation is taken into account); that both the typicality and the similarity of voice features get analysed, and language is seen as a "moving target", and not as a stagnant, stative thing (something we all ought to remember, especially for pronunciation!)

The work that people in FSS do is quite tricky, as it requires objectivity, conservatism and caution bearing in mind that their duty is to the court, and the results can only answer specific questions but clearly not assess levels of innocence. Human analysts can and do fail, and problems like confirmation bias and perceptual priming are to be avoided. We were shown a few cases where experts cannot agree on what the words of a certain segment are, and different kinds of evidence are weighed in order to submit candidate hearings. Plus, as we were told and show, the work that FS scientists do differs greatly from what you can see in CSI and other crime series!

Some interesting things about FSS that Watt has mentioned include:

• the need to focus on the individual as a potentially unique language user, and of each human voice as a holistic Gestalt as well as a constellation of features;
• the fact that even though the degrees of freedom in individual speech is unknown, the assumption is that there is  less intraspeaker variability than interspeaker variability;
• the quest in FSS is to identify most potent/reliable speaker discriminants;
• the fact that technology has advanced, but it will never yet replace human work.

Some of the tasks that FSS carry out include: 

• speaker profiling
• transcription
• resolution of disputed content
• speaker attribution
• enhancement
• authentication
• design of voice parades
• acoustic scene reconstruction
• language analysis for determination of origin (LADO)
• speaker comparison 

The University of York has an MSc in Forensic Speech Science, and if you want to read further about the programe, just click here.