This is a short primer on some important concepts regarding experimental design (especially as it is used in ecology) that will likely be useful while reading primary literature this semester.
Ecologists take multiple approaches to study the natural world. The goal of an experiment or a study is to learn more about scientific hypotheses. A scientific hypothesis is a proposal about the relationship between things in the natural world. For example, a relationship may exist between amount of sunlight received by an area of ground and the number of grass stems in that area.
There are three main types of experimental approaches used in research:
- Survey (“natural experiment”)
- Manipulative experiments
- Modeling experiments
Studies that have been published in peer-reviewed literature include one or more of these experimental approaches. Each of these experimental approaches has strengths and weaknesses and whether one is better than the others depends on the goal of the study. Review papers (such as Thompson 1998) also appear in peer-reviewed literature, and generally summarize all past and contemporary research on a specific topic. However, when writing your grant proposal, we do not want you to write a review paper.
Following is a little more on each of the experimental approaches.
Survey or Natural Experiments
- looks at variables as they change in unmanipulated circumstances
- "unmanipulated" here means that the researcher did not change things. Many studies are done following large natural events (ie: storms), or to gauge the effect of natural environmental variations (ie: change in elevation).
- example: Jordan wants to measure whether toxins from a paper mill affect crayfish. He counts the number of crayfish at varying distances from the mill.
- pros: clearly actually occurring in nature
- cons: hard to control other variables – to control for variability, many studies span 5-30 years, have large areas, many samples, etc; difficult to elucidate mechanisms behind patterns
Manipulative experiments
- purposefully changes one or more variables and measuring the outcomes
- may be conducted at multiple scales/settings such as in a lab, a greenhouse, or in setups created out in the field
- example: Jordan wants to measure whether toxins from a paper mill affect crayfish.
a) (Lab experiment) He collects water at varying distances from the mill and takes it back to his lab. Then, using this water in separate tanks, he tries to grow crayfish. He compares the growth of crayfish in the different tanks.
b) (Field experiment) He takes baby crayfish from the same brood and places them at varying distances from the paper mill. He measures the growth of each crayfish through time. (He has super crayfish-tracking ability).
- pros: much greater control! can isolate specific variables and control them; may be better at determining mechanisms
- cons: not natural settings/conditions – what you have measured may actually never occur in nature; set up and maintenance of experiments may be costly
Modeling Experiments
- construct a model based on scientific theories or experimentally collected data
- often then compared with experiments to gauge model accuracy
- ex: Jordan wants to measure whether toxins from a paper mill affect crayfish. He makes a model of how he thinks crayfish populations should change through time when toxins are present. He then measures crayfish populations.
- pros: generally require fewer resources; have greater predictive abilities (useful for conservation/management decisions or advising);
- cons: all models are wrong, some models are useful
© Asya Rahlin and Harmony Lu 2011
