Tuesday 4 November 2014

Machines can hear it

More from the No Such Thing As A Fish podcast. The Elves have an incredible breadth of knowledge, but by necessity cannot have an in-depth understanding of all the facts they present. This becomes obvious when they talk about something that one happens to know something about. This morning, I was listening to them discussing accents, and they said a number of things that reminded me that there's a curious mismatch in terms of how much people like to talk about language and how poorly equipped they are to do so. (Linguistics is not the only subject to suffer from this, of course. Psychology springs to mind.)

The most striking thing was their discussion of (by now fairly old) research analysing the Queen's Christmas speeches over the years, which concluded that the Queen does not speak the way she spoke when she was younger. The reason that this is interesting is that it shows that older people's language shifts, not just young people's. But the shifts documented in that study are slight, and affect vowels, which are notoriously hard to pin down.

Because we need to be able to tell the difference between bet and bat, we hear vowels as categorically different sounds, but they are actually points in a continuous potential noise-range (you can easily slide from e to a with no break in between, and the intervening sounds are halfway between the two vowels). What distinguishes vowels from each other is their relative 'distance' from each other (scare quotes to indicate metaphorical usage and also technical terminology - the distance is literal as well).

This means that when sounds change very slightly, you might not be able to hear it, or you might not hear it reliably because you're expecting to hear a particular sound (the McGurk effect is a famous demonstration of just how useless you are at hearing a sound if you're expecting to hear another). If you can't trust your own ears, you aren't going to do a very convincing scientific experiment, are you? So people use specialist equipment to measure sounds, and then they can analyse these measurements to determine exactly what sounds were produced and how (for instance) the Queen's vowels have changed over time. So, one change they noticed was that her vowel in words like had was produced lower in the mouth in the 1980s than in the 1950s, causing it to sound a bit more like had and a bit less like hed.

The Elves reported the study and in the course of the conversation, said 'You can't hear it though. Machines can hear it'. This was met with astonishment from the audience and the other Elves, but it's no stranger than measuring anything else with a machine (blood pressure, brain activity, blood sugar levels, gravity, atmospheric pressure, wind speed, radioactivity...). It's a peculiar quirk of language that because we all do it with a reasonable degree of consciousness, we think we must know all about it, even though we don't know how our lungs or digestion work without explicit teaching.

So I think the moral of this is that everyone should come to my department's forthcoming event, consisting of an exhibition, two film screenings and a public lecture, and learn more about how amazing language is!

No comments:

Post a Comment