In this video I explain how signal detection theory relates to psychophysics and the study of absolute and difference thresholds. I also explain how response criteria play a role in signal detection theory and the possibility of type I and type II errors. Finally, I consider applications of signal detection theory in daily life, from detecting dangers to dating.
Don’t forget to subscribe to the channel to see future videos! Have questions or topics you’d like to see covered in a future video? Let me know by commenting or sending me an email!
Need more explanation? Check out my full psychology guide: Master Introductory Psychology: http://amzn.to/2eTqm5s
Video Transcript:
Hi, I’m Michael Corayer and this is Psych Exam Review. In this video we’re going to look at signal detection theory. So in the last video we talked about psychophysics and we talked about these ideas of thresholds.
We had the absolute threshold, which was the minimum amount of stimulation that’s needed for us to detect a stimulus, and then we talked about the just noticeable difference, or the difference threshold, and this was the amount of change that’s necessary in a stimulus in order for us to detect that a change occurred.
So one of the problems that we have in both of these situations is the problem of “noise”. So what do we mean by “noise”? Well a noise is anything that can interfere with our detection of a stimulus. So it’s not specific to hearing, it’s anything that interferes.
So we use this term noise to refer to anything that interferes with detection. What you’ll realize is that noise is everywhere, for all of our senses. The world is noisy. If you think about trying to measure your absolute threshold for hearing you can imagine that I want to find the quietest sound that you can hear well I’m going to have other noise that gets in the way.
Even if I put you in a soundproof room we’re still going to have a problem of noise, right? Because there’s no such thing as a perfectly quiet environment. As soon as you get into that quiet room you’re going to realize that there’s air in the room and there’s air molecules moving around and some of those are going to hit your eardrums and that may be interfering with your ability to hear the noise. Or you’re going to realize that maybe there’s some ringing in your ears that you don’t always notice, but you notice it now that you’re in a quiet room or you become aware of your heart beating, blood moving in your body. That might interfere with your ability to hear. If your heart beats of the same moment as the sound that I’m playing that you’re trying to detect it might cover the sound up. You might not be able to hear it anymore. Or depending on how you breathe, your breathing is going to make noise. All these things are going to interfere with your ability to detect a stimulus.
This is why trying to find absolute thresholds is actually really difficult. There’s all different types of noise that can interfere. For vision, for instance, I could put you in a perfectly dark room but you actually still have some things that you’re going to see. You’ve probably experienced this if you’re in total darkness, it’s not just pure blackness that you see. You see sort of little bits of light and there’s spontaneous neurons firing in your brain. You’re sort of almost hallucinating these little things that you think you’re seeing.
That’s going to interfere with your ability to detect light that’s actually in the room. So we have this problem of noise and we can look at, in signal detection theory, what we look at is the possible outcomes. So we have your response, which is if we use the hearing example, let’s say you put on some headphones, you’re in a quiet room and I play a sound and you either tell me yes you can hear the sound or no, you don’t hear anything. So those are your two responses, we’ll say you have to answer yes or no.
Well there’s also two possibilities on the side of the signal. Is there a signal, yes or no? Maybe I didn’t play a sound. So over here we have the presence of a signal, yes or no, and we have your response, yes I hear it, no I don’t. This gives us this matrix of four possible outcomes.
So the first possible outcome is I play the sound and you say “I heard that”. We call that a hit. Another thing that’s possible is I play the sound but you don’t say that you heard it. And that would be a miss. There’s a signal present but you don’t detect it. The other thing that could happen is I don’t play sound and you say “I heard that”. But there actually wasn’t a sound that you were supposed to hear. So in this case, we call that a false alarm.
Of if you take a statistics course you may see this type of error called a type-I error. This idea that that there is not a signal and you think that there is. That’s a type-I error. Now there’s also the possibility that there’s no signal and you don’t say that you hear it. So I don’t play a sound and you say “I didn’t hear anything” and that would be a correct rejection.
So we have these two possible errors here, the false alarm which I mentioned is sometimes called a type-I error, and then the miss here where there is a signal but you don’t detect it. And this is sometimes referred to as a type-II error. Then you have these two correct outcomes where you have a hit, where there is a signal and you detect it, or a correct rejection where there’s no signal and you don’t think that there’s a signal.
OK so we can imagine this hearing test here and these are the four possible outcomes. But you might realize well sometimes it’s really easy to decide if I should say yes. “I definitely heard that” or “I definitely didn’t hear anything”.
But there’s a lot cases, because we’re trying to find this minimum stimulation, right, so it’s going to be really hard to decide if you heard it or not, right? Because also we have this problem of noise. So a lot of time you’re going to be sort of in the middle. The real question here is, do you lean towards saying yes are do you lean towards saying no?
This is what’s known as your response criteria. So your response criteria is essentially asking “what do you do when you’re unsure?”. Do you lean towards yes or do you lean towards no. So if we say you have a liberal response criteria that means you lean towards saying yes. So you think you heard it, so you say “yeah, I heard that”.
Or we can say you have a more conservative response criteria and that means when you’re unsure you lean towards no. You want to be really certain that you heard it, so if you’re uncertain you say no. Now we can think about how this is going to influence your overall pattern of responding. So you could imagine two people, imagine I take this test and you take this test. We could imagine that we have exactly the same ears, that the stimulation, our ears are working equally well, they’re identical.
But if we have different response criteria, if I lean towards saying yes when I’m unsure and you lean towards saying no when you’re unsure, our answers are going to end up being totally different even though our stimulation is identical. Even though our ears are working exactly the same way. This is a big problem in signal detection theory. We have this question of how liberal or conservative should our response criteria be? The same ears could give us different answers and we might have to think about, “well, which of these errors is going to be worse?”. Do I want more false alarms? If I lean towards yes I’m going to get all the hits but I’m going to have more false alarms. If I lean towards no I’m going to have more correct rejections but I’m also going to have more misses. I might have to think about which of those is better.
This brings us to applications of signal detection theory. So it’s not just for absolute thresholds or just noticeable differences. What I hope you realize is that signal detection theory applies to almost every area of your life. Because there’s a lot of decisions you have to make and very often you don’t know should I say yes should I say no. It’s unclear what the correct answer is. The truth is that life is “noisy”. Most of the signals we’re trying to detect are hard to detect and there’s a lot of things that interfere.
So this can be things like wondering, a police officer wondering if someone is reaching for their wallet or for a weapon. This is a life-altering decision that you might have to make. You have to think about which type of error is better or worse. Or if you’re on a jury and you’re thinking about, you know is this person guilty or innocent? Well is it worse for me to have an innocent person falsely imprisoned or a guilty person go free?
Which type of error do I want to lean towards? So the same people given the same information could come to different conclusions. You could imagine a doctor looking at the results of an exam. Looking at a scan and wondering is this noise in the data or is this a tumor? How liberal or conservative should I be? Should I recommend preventative surgery or is that unnecessary?
Or if I don’t recommend it, if I lean towards saying it’s probably nothing, it’s probably noise, then six months later it turns out it was a tumor and now it’s grown. These are hard decisions that people have to make.
You could think about other decisions when you think about which college you’re going to choose, which college is right for you? it’s hard to make that choice. Or in dating you might think about finding the one, the person you want to marry. When it comes to dating, how liberal or conservative should you be on accepting that first date?
On the one hand if you’re too conservative, if you always lean towards no, you’ll correctly reject all those people that aren’t right for you but you might miss the person who would be really good. On the other hand if you lean towards being too liberal and you say yes to every first date, you’re more likely to find that one person to get that hit, that person who’s right for you, but in the meantime you’re going to have a lot of false alarms. You’re going to go on a lot of dates that you’re wasting your time on.
So I hope you see that we can really apply signal detection theory to all sorts of decisions that we make that I encourage you to think about all the different areas where signal detection theory can be applied to your own life. And think about what’s my response criteria? Which type of error is better or worse when I’m making this decision?
OK so I hope you found this helpful, if so, please like the video and subscribe to the channel for more.
Thanks for watching!
