# The Scientific Method

### Observation

1. Inductive reasoning (from the particular to the general) to generate a testable hypothesis
2. Deductive reasoning (from the general to the particular) to generate predictions about future observations
3. Tests of the predictions with controlled experiments or observations
4. Logically interpret the results – conclusions
5. Start over again with the new observations added to the old observations

### Hypotheses

• Can never be proven true, only disproven
• Any set of observations can be explained by many hypotheses
• Ocham's razor (the explanation that requires the fewest assumptions is preferred)
• Must explain all known observations
• Must be testable
• Must be based on logic and natural processes
• Hypotheses that have withstood many tests and explain many observations are called theories

# An Example

### Observations

• Marsh grasses in different locations grow at different rates and to different heights.

### Questions

• What causes the difference in growth rate and height?
• Inductive Generalization
• A large number of observations on growth in marsh grasses and other plants suggest some specific questions
• Amount of nutrient – Nitrogen (N), Phosphorous (P) or Potassium (K)?
• Tidal flow or depth?
• Activity of mud-dwelling animals?

### Hypothesis

#### A postulated answer to the question

• An increase in N levels causes an increase in growth and height
• Null form: an increase in N levels has no effect on growth
• A null hypothesis is the hypothesis that something is not the answer – the null is very useful because while you can not prove anything true, you can prove things false
• Hypotheses lead to deductive predictions
• If the growth and height of marsh grasses is increased by increased N levels, then grasses with added N should have greater growth and height.
• Deductive arguments are tested. (by Experiment or Observation)

### Experiment

• How to test the hypothesis - What data - How many repeated tests - etc.

• Then add Nitrogen to one of the plots only (randomly chosen, of course) and measure the effect of the added Nitrogen on the growth of the grass. If the hypothesis is true then the plot with N added will show more growth

• Experimental Variables:
• Independent: to be varied by investigator = N level
• Dependent: growth rate, height
• Controlled Variables:
• many - water depth; soil organisms; salinity; other nutrients; etc.
• try to assure that paired plots are as much alike as possible

## Experimental Design

### Preferred test is a controlled experiment

• Only one independent variable is changed at a time
• All other independent variables are held constant (controlled)
• usually can not control all independent variables so control all variables that you believe are relevant
• Must be reproducible
• this checks for other unknown variables and stochastic effects
• Usually have an alternate hypothesis and a null hypothesis
• try to prove that one or the other is false
• can only have statistical proof (95%)

### When using humans you need to control for bias

• Double-blind randomized trials
• experimental subjects are assigned to control and experimental groups randomly
• use a placebo for the controls
• subjects do not know whether they recieve the treatment or the placebo
• researcher does not know whether they recieve the treatment or the placebo until after the results have been analysed

### If a controlled experiment is impossible, then use controlled observations

• Some examples
• big things: planets, stars, etc.
• things that take a long time: evolution, geology, etc.
• unique events: origin of life, the universe, earthquakes, etc.
• expensive or dangerous events: toxic waste spill, nuclear melt down, etc.
• Try to control for all relevant variables
• Must predict the result of new observations

### Assignment due before next class meeting:

After reading the class notes on designing experiments, think about some example of pseudoscience or alternative science (astrology, palm-reading, accupuncture, intelligent design, echinacea, etc.) you've heard about, come up with a scientific hypothesis based on the idea and then design an experiment that could be done to test the hypothesis. Your experiment should have a clear question, a hypothesis, a prediction, and an experimental design that would test your prediction.

This document is maintained by: Jeff Bell
Last Update: Tuesday, July 11, 2006