Psychology 120.3 Lecture One September 3 2025
The Ten Big Topics
students are encouraged to video this lecture for additional information
Topic One: How Do We Know When We Are Doing Science?
The first step in creating a working scientific theory is to create the null hypothesis. This process is counter-intuitive. We naturally want to prove something to be true, but the first step in the philosophy of science is to declare that Phenomenon A has no connection to Phenomenon B.
Let's use an example that we could try out in the fitness centre. I want to see if there is a measurable relationship between the use of a cross-country training machine and fat loss. Note I am not interested in overall weight loss.
We can use CC for cross country, and FL for fat loss.
Topic Two: How Do We Design Hypotheses?
Here is the null hypothesis, symbolized as H0.
CC ○ FL
Here is the first alternate hypothesis, that using a cross country trainer will cause fat loss.
CC --> FL
Note the use of the word 'cause'. There are many suppositions underlying that word. In neuroscience and psychology in general, we rarely claim that one condition causes the other. Instead we use correlations, that the two variables--Cross-Country and Fat Loss are connected.
CC --><-- FL
Topic Three: What Is A Correlation?
A Correlation is defined as the degree to which two variables--in this case cross-country trainers and fat loss-- tend to move together in either the same direction (positive correlation) or opposite directions (negative correlation). A correlation does not mean that one variable causes the other; it simply shows that there is a link between them. A correlation coefficient, a numerical value between -1 and +1, quantifies the strength and direction of this linear relationship.
https://www.scribbr.com/statistics/correlation-coefficient/
Here is where the problem in thinking begins. If a subject does twice as much training as a subject in another group we would expect fat loss to increase. But fat loss is a negative process, with a negative result. After all we call it 'losing weight'. So our goal in the experiment is to demonstrate a negative correlation, not a positive one.
Topic Four: A Practical Example Of How To Use Correlations.
So..let's create a practical example. One subject--an 18-year old female--wishes to go from a weight of 80 kg (≈ 160 lbs) to 54 kg (≈ 120 lbs). If the correlation was -1 there would be a fat loss at the end of every cross-country session, measureable with calipers on her upper arms. If the correlation was +1 she would gain bodyfat at the end of each session. In reality, a very good result would be -.8, a mediocre one -.5, a low correlation of -.3.
Topic Five: Why Is A Negative Correlation So Hard To Understand?
So, in conclusion correlation coefficient is a statistical measure (ranging from -1 to +1) that . A value of +1 indicates a perfect positive linear relationship, meaning one variable increases as the other increases. A value of -1 indicates a perfect negative linear relationship, where one variable increases as the other decreases. A 0 suggests no linear relationship, but the variables might still be related in a non-linear way.
Think it through for yourself. What underlying variables were ignored? These have a lot of names, but for this case we will focus on a third hidden or latent variable. The first one that comes to mind is diet. Try to think of at least two more of these hidden or latent variables yourself.
So, how do we design such an experiment that will yield measurable, and unbiased results? Those calipers don't lie. We go back to the beginning and create our null and alternate hypotheses.
Null Hypothesis: There is no connection between cross-country sessions and fat loss.
Alternate Hypothesis One: As the number of cross-country sessions go up, percentage bodyfat goes down. That kind of hypothesis is called causative.
Alternate Hypothesis Two: As the number of cross-country sessions go up, percentage bodyfat goes down, due to a hidden or latent variable. That kind of hypothesis is correlative.
Topic Six: What Is An Experiment?
Now things get difficult. We are forced to apply Newtonian 18th Century Physics to Psychology, which has been a hit-and-miss approach for over a century. The experiment has to be whittled down to just two variables: cross-country training and percentage bodyfat. We assume that there are not hidden (or) latent variables to affect the results of the study. In this case we will assign the cross-country training to the independent variable, the one we control. Percentage bodyfat will be assigned to the dependent variable, the one that changes due to the action of the independent variable.
Topic Seven: Control and Experimental Groups.
In its purist form the subjects of the experiment will be randomly chosen to either take part in the experiment as part of a control group (subjects who do not use the cross-country trainer), or as part of the experimental group, the subjects who do use the cross-country trainer, according to the plan created by the researchers. In this case, we will use ten training sessions of one half-hour, each session occurring every other day over a two-week period.
Topic Eight: Subject Variables and Demand Characteristics.
Note that our first classical experiment was closer to kinesiology than psychology. That was not an accident, as you discover when we dive in the neuroscience. Subject variables such as age, diet and lifestyle will dramatically change the outcome of the experiment, and we can't do much about that, since we are focused on only the two variables. What can we do to minimize this, since we supposedly need to randomly assign subjects to either the control or the experimental group (according to classical design).
Demand characteristics have been minimized in this design because we don't need to create a questionnaire to ask subject opinions, though this would have to happen in a peer-reviewed experiment. Subject will often tell the researcher what they want to hear, rather than the truth.
Topic Nine: Matched Sampling (Samples and Pairs)
What we can do is to create a quasi-classical experiment by using matched sampling. In this case, subjects are paired based on shared characteristics other than the variable being studied, such as age or gender. This reduces the effects of those hidden, latent variables. Instead of comparing two unrelated groups, in this case a matched sampling design will collect data on paired subjects. Practically speaking, it would mean placing an 18-year old woman athlete in both the control and experimental groups. This then would complete the basics for our before-and-after experiment.
Topic Ten: Reliability and Validity.
And now we come to the hardest part of all, showing that the experiment was worth the time and effort. In this case, reliablity is fairly easy because we have fine control over the cross-country trainer. Most modern equipment generates the data, which can be input directly into statistical software such as SPSS (Statistical Package for the Social Sciences). The same goes for the percentage bodyfat measurements.
So, the experiment is high in reliability. How about validity? This splits into two types: construct and concept. Construct validity deals with how well the experiment was run. For example, did all subjects complete the full half-hour at the assigned performance level? Let's assume that the cross-country trainers were set to moderate hills and moderate speeds; did anyone change the settings, and the researchers did not find out until they tabulated the data? Things like that happen. If not, then the experiment is high in construct validity. What was claimed to be measured, did get measured.
Concept validity is much harder to demonstrate. Is a drop in percentage bodyfat as measured by the calipers a true measure of fat loss, or are there hidden variables? Is a cross-country ski machine the best way to drop bodyfat, or is it actually a moderate improvement that will fade away after the two week training program? At this point motivational psychology kicks in.
Check out YouTube videos for 'biggest losers' to see where these questions lead.