Hi, Andy Smith.

Today in a study of experimental research methods in psychology,

we're going to look at variables.

Now the dictionary defines a variable as an element that's liable to vary or

change, and in the experiment there are lots of things that change.

Some that we want to change, some that we don't want to change,

some that we have control over, and some that we don't have control over.

There are many variables, many things that can change in an experiment.

Now the first is the dependent measure.

That's what we are measuring, the dependent variable.

What we're measuring the experiment, what is observed, and in psychology that's

always behavior, and we don't know whether that's going to change or not.

That's the empirical question, will we see difference,

will we see change in that dependent variable, that behavior?

There's also the independent variable, and that's what we manipulate, so

we control that change as the experimenter.

We manipulate something in the experiment.

We want to see whether that produces change in the dependent variable

or behavior.

So what we're trying to do is to show that the dependent variable

behavior is a function of the independent variable.

We want to show in the experiment that the dependent variable

is controlled by the independent variable, and

that is the empirical question that we're asking, does this happen in an experiment?

Now independent variables used in the experiment, Are controlled,

they are manipulated, but

we have to make sure that that's the only thing being manipulated.

They really can't be in the experimental, what they call confounded variables,

the third kind of variable, and confounded variables means that something that's

changing that we don't really want to change because it can mess up

the relationship we're trying to draw between the independent and the dependent

variable, and there are really three different kinds of ways this can happen.

The first is the third variable problem.

That's a variable that's confounded with the independent variable that we're trying

to manipulate.

There's a placebo-like effect, where an independent variable might

be producing effects that we don't really want to happen, placebo,

we can control for that, and then the third is mediating variables.

Maybe the independent variable that we're manipulating is not the one that's

controlling the dependent variable that we want to measure, but something in between,

that's causing that to happen.

We'll look at the examples of these three different kinds of confounding variables.

The first is the third variable problem, and that's the true confounded variable.

That is, we might have an experiment where, let's give an example,

like we manipulate plane violent video games, I'm making this up.

That's an independent variable.

Playing video games.

So that's when we manipulate that.

We have kids that play violent video games, and the kids who do not play video

games, and we measure aggressive behavior on the playground.

That's a dependent measure, and we find

a causal effect between the independent variable and the dependent variable.

The more the kids play violent video games, the more they engage

in aggressive behavior, touching of other people, on the playground during recess.

So we have a causal relationship that we have between playing violent

video games and aggressive behavior, at least we think we do.

But, what in fact, there's a third variable that

actually controls both of these variables?

Like, for example, a personality variable which might be how willing are you to take

risk, and you might argue that personality risk taking

actually will produce the tendency to play violent video games, but

also aggressive behavior on the playground.

So, there's a third variable that's confounded with

the experiment that really produces the change that we see, but

not a change that we want to necessarily, because they're correlated,

not a change that we necessarily want to draw a conclusion about.

So a third variable could be an important confounding to the experiment.

Now, the second confounding variable is a placebo-like problem.

You think of placebos when you think of taking drugs.

I have a drug that I think cures headaches, I give the drug to one group,

I don't give the drug to another group,

the people that took the drug have reduced headaches.

Then I say, okay, my drug works, but the problem is,

research simply shows that intervening in anyway will produce change.

So just the fact that I gave you a drug and said this is going to help you,

and you expect it to help you, it most likely will help you, whether it does or

not biologically.

That's called a placebo.

We have a placebo-like problem.

So playing video games, might be producing the aggression on the playground.

It doesn't matter whether video games are violent or non-violent.

So we have to make the comparison between that makes that

comparison a true one taking care of the placebo effect.

Just like how I would have to give you a placebo, a sugar pill if I really want to

test my drug, and it is effectiveness against headaches, because the headache

improvement has to be over and above what the placebo drug would happen.

So we know that playing violent video games,

when I measure that, produces more aggressive behavior in the playground, but

I might want to also have a condition that playing video games,

but not playing violent video games as the placebo.

And so that would be playing non-violent video games, and

that would need to produce significantly less aggressive behavior

than playing the violent video games in order to make the conclusion

about violence in the video games and aggressive behavior on the playgrounds.

You have to solve the placebo-like problem.

In the third confounding variable are called mediating effects, and

sometimes these are good to find,

because they tell us even more information about the situation that we're studying.

But let's say, for

example in the experiment, manipulated playing violent video games and

measured aggressive behavior on the playground, we found the causal effect,

just like I said, between the playing the non-violent video games and the need for

contact others in the playground.

But what if the aggressive behavior is just that,

it's the need of contact with others?

It's not necessarily the aggression, but simply the need to contact others

that's produced by playing the violent video games.

So it's a mediating effect between playing the violent video games, and

what we see on the playground.

It produces something which mitigates in between the two variables in

our experiment, the variable we manipulate and the variable that we measure, and

that's called a mediating effect, and in fact, if we find mediating effects, and

understand them, they can really tell us a great deal about what's going on

in this situation that we're studying.

So we have to at least understand though there might be mediation,

of that relationship, by another variable that we're not measuring.

Now let me give you an example from my own research.

I study aging and memory, and like many others, I have found in many, many

different kinds of situations, that memory declines with normal healthy adult aging.

I'm not saying Alzheimer's disease, I'm not saying pathology,

just a normal aging, starting at about age 25,

we see declines in the ability to recall things from memory, just due to aging.

So we're saying that normal adult aging shows us decreases in what we

call episodic memory, remembering something from our past experience.

So remembering something that we could look up in our internal diary,

like where did I leave my keys?

When was the last time I saw my brother?

These are episodic things, and

those ability to remember those, declines in normal aging.

There are others that have argued that it's not aging that's reducing the memory

declines, but health.

We know that ageing is associated with decreased health.

The greater probability of having problems with health is an age related finding.

Health declines with aging, and maybe that's what's producing the memory change.

So age affects health, and health is what produces the changes in memory,

because we know, when we're not healthy,

we're not able to be cognitively as alert as we were when we are healthy.

So, normal adult aging produces decreases in health, mediating,

and that produces the change in the decrease in memory.

So that's a possible mediation effect, and I actually studied this,

whether age affects memory directly or whether age affects memory indirectly

through changes in health, in an experiment that was done many years ago,

and this is the way we did it.

It's really Earles, Connor, and Smith Park, 1997, and we asked the question,

whether health was a mediating effect with the aging memory relationship?

Now here's a venn diagram, I'm just showing okay there's this age memory

relationship, and that's in the dark blue between the two.

There's a large overlap between aging and memory.

Aging produces changes in memory.

When we added health to the equation, we found that health accounts for

some of this relationship, but really a very small part of it.

A very small part of the total relationship between memory and aging.

So, in this case we found that health is not a very good mediating effect.

It's that aging itself, what happens in the brain,

with aging is really what produces the changes we see with memory, and

we know a lot about that now, about the hippocampus, and

how it's really important for formation of new memories, of episodic memories,

and how the hippocampus shows changes with normal aging.

Just to show that we actually did a study to look at something that

might mediate between age and memory, and what we did was come up with

a measure that really just look at neural efficiency of the brain.

An example would be perceptual speed,

how fast we can make decisions about perceptual things happening, and

look to see whether perceptual speed mediated the age memory relationships,

and sense it measures brain efficiency, we really expected that to happen.

So in a different thing, here's the age, memory relationship again,

unlike health when we added perceptual speed it accounts for

most of the relationship between age and memory.

So now we have a mediating effect, and we understand better what it

is about the age, memory relationships in terms of causation.

So sometimes mediation can actually expand their knowledge.

We want to make sure that we,

if we have mediation we don't expect it that we understand it.

So what are the variables, we have dependent variables, what we measure,

we have independent variables, what we manipulate, and we may have confounding

variables, things that covariate with the the independent variable.

Third variable problems, or it might just be a placebo-like effect,

because we're doing anything with the participance in that research.

We gotta make sure that anything can be controlled for by looking at what

is the actual independent variable we want to measure, and in my example,

it was whether or not the video games were violent versus non-violent.

In that we made a very much more precise independent variable we've controlled for

the placebo effect of just simply watching video games, and

mediating variables, variables that can mediate in between the independent

variable that we're manipulating, and the dependent variable that we're measuring.

Thank you.

Lesson 1B: Variables

Experimental Research Methods in Psychology

Meet the Instructors