CSUF Department of Psychology
Unit Banner

NOISE & EXTRANEOUS VARIABLES

“Noise” isn’t a term you’ll find in most Research Methods textbooks, but it seems to be a good term to describe effects in research that might disguise what you’re really looking for. That is, anything that clouds or obscures your measurement could be considered "noise." Take, for example, this picture:

 

 

 

 

Now place your mouse on the picture to remove the noise. [Note: If you're using Explorer, you might get a message that 'Active Content' needs to run on your computer. Click on the bar with that message and allow Active Content to run.]

Initially, the picture is obscured by the noise -- it kind of looks like a face but you can't really tell. What we want to do, as researchers, is to get a "clean" measure (or, in scientific terms, a valid measure).

Let's take an example. If you're trying to measure whether people are happier when the weather is sunny, what kinds of things do you think would affect your measurement of happiness? What if someone just took a really hard test? Might this affect their happiness, independent of the weather? What if their cousin just had a baby? What if they found a sweet parking spot? What if they just got 50-Cent's latest CD from a friend? What if they just got a ticket for speeding? What if...what if...what if...?

The point here is that a researcher wants to try to eliminate or control the "noise" so that the measure is what you're really trying to measure, rather than something else. A valid measure is one in which you measure what you're trying to measure. Usually, this means that the noise has been reduced. (By the way, unlike the picture above, when you're looking for something psychological, you can NEVER remove all the noise; human behavior is simply much too complex.)

There are many contributions to ‘noise’ (aka experimental error) in any psychology study. These contributions come from extraneous variables. What we try to do in research is a) identify the sources of noise, and b) do what we can to eliminate those sources (or, in research terms, control the extraneous variables). We’ll start out identifying the sources; later you’ll learn about the concept of control.

Let’s say I want to know if holding a pencil in your teeth makes people happier (on average). I choose you and another classmate (Sue) to be in my “no pencil” group. (Oh…and by the way, the classroom is *really* hot.) I then tell you both to "really think about your mood." On a scale of 1-7, where 7 is 'Extremely Happy' your ratings are:

You
Sue
6
2

Why are you and Sue different? Maybe you're just happier than Sue. Maybe she just received bad news. Maybe she doesn't care about this silly experiment. These are all sources of noise from extraneous variables. (We'll discuss types of extraneous variables later.)

I then choose two Marx brothers, Zeppo and Harpo, to be in my ‘pencil-in-teeth’ group. [Note: The Marx brothers were comedians in the early 20th century.] They’re in a comfortable climate-controlled room. I tell Zeppo and Harpo to write the first rating that comes to mind. Here are their data:

Zeppo
Harpo
7
5

Again, why are their ratings different? Because of extraneous variables.

Now let's calculate means for each of my groups:

No Pencil: (6 + 2) / 2 = 4
Pencil-in-Teeth: (7 + 5) / 2 = 6

Clearly, 6 is greater than 4. Now…can I say that holding a pencil in one’s teeth makes people (or…CAUSES people) to be happy? Think about reasons why you agree or disagree.

Can I make an inference about my data with only two people in each group? What about having comedians in one group but not the other? What about my instructions? Could that have made a difference in this experiment? What about the temperature of the room?

======================

You probably had no problem coming up with answers to these questions, and all of your answers probably related to the idea of NOISE in any psychological study. Noise is experimental error from extraneous variables, something that might have affected your data that you don’t want in your study (but as you'll find out...you can't get rid of ALL extraneous variables).

Extraneous Variables

Any variable that is not manipulated (i.e., not an independent variable) but still might have an effect on the dependent variable. Extraneous variables are the cause of experimental error (or NOISE), making it more difficult to measure the relationship between variables or interpret the results of an experiment (i.e., the effect of an independent variable on a dependent variable). There are four classes of Extraneous Variables:

Participant (or Subject or Person) Variables

A type of extraneous variable in which one participant is different from another such as intelligence, motivation, expectations, athleticism, mood etc.

  • Maybe Sue gave a low mood rating because she got a parking ticket that morning, or maybe her roommate drank all the milk, or maybe she got a virus on her computer, or maybe Sue hates me (the experimenter), or whatever.
  • Maybe you had a high mood rating because the Starbuck’s worker gave you a free coffee (yeah…right).
  • Maybe Zeppo is just a happier guy.
  • Maybe Zeppo’s “7” on the mood scale is Harpo’s “5.”

Treatment Variables

A type of extraneous variable which arises from the way participants are assigned to different treatments or conditions. This can include the ORDER in which tasks are taken in an experiment.

  • Maybe the ‘pencil in teeth’ group was happier simply because the comedians were in that group.

          Task Variables

A type of extraneous variable which arises from the procedures (e.g., difficulty of the task, time allowed for completion, or the nature of the instructions).

  • The ‘pencil-in-teeth’ group was told to make a gut-response to their mood whereas your ‘no pencil’ group was told to really think about it. Maybe the more you think about your mood the lower it becomes…maybe. (Just another hypothesis, just like the other explanations. (Remember, doing research is really about eliminating alternative explanations.)

     Situation Variables

A type of extraneous variable that involve the environment of the study (e.g., physical characteristics of the space like temperature, noise level, demeanor of the experimenter, presence of other people).

  • Maybe the hot, uncomfortable room lowered your mood versus the comfortable room of Zeppo and Harpo’s.

As you can see, extraneous variables may have an effect on your measurement that might disguise what you're really trying to measure. When you measure any variable, there will ALWAYS be effects of extraneous variables. Even in the most controlled environments (e.g., a small dark windowless room with no decorations), there will be differences in your participants, there will differences in the mood of the experimenter(s), etc., that may affect what you're trying to measure.

By the way, having only four participants is not nearly enough to say anything about how having a pencil in your teeth might affect mood. This will be discussed further at another time. However, you should know this pencil-in-teeth thing is an actual finding that has been replicated over and over (therefore, it has RELIABILITY). That is, for most people most of the time, holding a pencil in their teeth, not touching their lips causes their mood to elevate. You can check out a non-scientific site on this: http://bipolar.about.com/cs/humor/a/000802_smile.htm