Thursday, January 19, 2012
Thoughts on Teaching Listening (Part 1)
I can’t remember who said it (I have a feeling it may have been Penny Ur), but I remember hearing a quote about teaching listening once that really made me stop and think:
We don’t really teach listening; we just keep testing it.
Whoever it was, I think he or she had a very valid point. Our standard methodology for teaching listening is a cycle of giving listening tasks and then asking questions in order to test the learners’ comprehension of what they have heard. In our defence, of course, it is difficult to see how we could do otherwise. Like reading, listening is a receptive skill that can only be developed through repeated practice, so there are good reasons for teaching it the way we do. Anyway, I was recently asked to do a presentation on this topic, and I started thinking about aspects of listening that do actually need to be taught rather than simply practiced. The first thing that came to mind was a list of general principles of which learners often seem to be unaware, and I want to write about the first of those today.
The first point is that we listen with our brains, not our ears.
Many learners think that improving their listening skills means developing their ability to recognize audio signals produced by proficient speakers of the language. In fact, the audio signal that most speakers produce does not contain sufficient information for it to be recognized in isolation. To hear an example of this, try saying the following sentence quickly.
“There was a great movie on TV last night.”
Now say it again at the same speed, but stop at “a.” Repeat the first three words, but keep your pronunciation exactly the way it was when you said the full sentence. Now dictate this segment of language to a proficient speaker of English and ask them to write down what you are saying. The chances are that they will be unable to do so. If you then complete the sentence, however, you will find that the listener is suddenly able to understand what just sounded like random noise a few seconds ago. The person will probably even tell you that they can “hear” the words now. Of course, the listener cannot hear “There was a” because you are not really saying it. The reason they think they can hear it is that after the words enter their ears, their brain takes over to decode the signal in the light of what went before, what came after, and what that person knows about the way English works. In other words, it is the brain that is doing the listening, not the ears.
You can see this process in action for yourself by playing around with an iPhone 4S. Even if you do not have one of these yourself, I am sure you have been bored to death by friends who do have one telling you about it and/or demonstrating its features. Anyway, Siri is a “virtual personal assistant” that, so its developers claim, is capable of recognizing natural speech. The interesting thing about Siri is that it only works when you have a good Internet connection. This is because it has to take the audio signal and process it through powerful servers in order to work out what has actually been said. If listening really were a matter of simply recognizing audio signals, this would not be necessary. The fact that it is necessary shows that it is the computers that are doing the listening, not the microphone.
I was quite skeptical of Siri before I tried it, because the voice recognition software I have used in the past has been worse than useless. Siri, however, is remarkably accurate, and it gets things right a lot more often than it gets them wrong. When it does make mistakes, however, it provides interesting insights into the challenges the human brain faces when it attempts to comprehend spoken language. One of the most difficult things about decoding the audio signal produced by spoken language is working out where the word boundaries lie. This is because different sequences of words can produce an identical audio signal. To give you an example of this, here is a true story about a message I tried to send the other day. On that day, I was feeling particularly pleased with myself because I had managed to get up at 6 a.m., and I decided to send an email to a friend to inform her of this remarkable achievement. Picking up my phone, I dictated to Siri, “I just broke the world record for getting up early.” When it had taken a moment to absorb the sequence of sounds that came from my mouth, Siri displayed the following message on the screen:
I just broke the world record forgetting up early.
If you think about it, the audio signal produced by my original sentence and the one produced by Siri’s transcription of it would be identical, but no speaker of English would interpret those sounds as Siri did because that sentence simply does not make any sense.
Another issue that both humans and computers have to cope with is the problem of homophones (different words that share the same pronunciation). One way software engineers attempt to deal with this is by taking into account the relative frequencies of words. For example, “rain” is more common than “reign,” so when in doubt, the computer will opt for the more common word. This, however, can lead to mistakes. As a follow-up to my first sentence, I continued my message: “Maybe I should contact the newspapers to tell them about my feat.” Before you read on, can you guess how Siri transcribed this sentence?
Of course, “Maybe I should contact the newspapers to tell them about my feet” is perfectly grammatical, and it would even make sense in some contexts. The problem is that in order to arrive at the correct interpretation, the listener has to hold in his or her mind a continually developing sense of what is being discussed. This sounds simple, and indeed it is—for humans. Unfortunately (or maybe fortunately!), however, it is still beyond the capabilities of even the most powerful modern computers.
To summarize, it is important for learners to understand that the reason they cannot “hear” clearly what proficient speakers of English are saying is that the speakers are not saying it clearly in the first place, so the sooner they give up on that idea, the easier it will be for them. Other proficient speakers can’t “hear” English in that sense either. What allows us to decode the signal and understand what is being said is our knowledge of English vocabulary and grammar, and our ability to keep track of topics over the course of a conversation. As I said, we listen with our brains, not our ears.