Design Your Own Experiment

Assignment

New-York-City-Central-Park-squirrel-lg
Squirrel in Central Park, New York City

For this assignment you will make observations, develop a testable hypothesis, and design an experiment to test your hypothesis. Begin by looking around your home (backyard, neighborhood, local park, etc.) for a species of animal that you can observe interacting with its environment such as squirrels, birds, insects, or lizards.

 

 

 

1. Observations

Once you have chosen a species you would like to study, observe its behaviors for a while. How does the species interact with its environment? How does it interact with other individuals of its own species, individuals of different species or with humans? Observe how it forages for food, how often it checks for predators, when it is most active, etc. It may be necessary to do some background research on your animal or system
Example from the marmot experiment: Li et al. noticed a difference in behavior between marmots that were exposed to human presence and marmots that were not.

Example from the mayfly experiment: Peckarsky et al. observed a difference in the size of mature mayflies (Baetis bicaudatus) in streams where brook trout were present versus streams where no brook trout were present.

2. Questions

After you have made initial observations of your animal’s behavior think of a few questions that interest you, based on your observations.

Example from the marmot experiment: To what extent do human disturbances affect the foraging and antipredator behaviors of marmots (Marmota flaviventris)? Are the effects of human disturbances different among juveniles, yearlings and adults?

Example from the mayfly experiment: Are the observed differences in size caused by the presence of brook trout and trout chemical cues?

3. Hypotheses

After you have thought of specific questions about your system, develop a hypothesis (a testable statement) that can be experimentally tested to answer your question(s).

Example from the marmot experiment: Motorized vehicles, pedestrians and bicycles will influence the behavior of marmots leading to, and juveniles, yearlings and adults may be affected differently.

Note: In the Li et al. study, researchers were open to the direction in which behavior changed. For example, marmots may exhibit depressed levels of foraging and elevated antipredator behaviors in areas of high human disturbance if humans are viewed as a threat. Alternately, realizing that humans do not pose a threat, marmots may exhibit elevated levels of foraging and depressed antipredator behaviors in areas of high human disturbance since humans may deter predators.

Example from the mayfly experiment: Brook trout chemical cues will affect the size of mature Baetis bicaudatus, leading to a reduction in the adult size and fecundity of mayflies emerging from streams where brook trout are present.

4. Experimental Design Part I: Define Variables

Once you have developed a hypothesis, design an experiment to test your hypothesis. Experiments have dependent and independent variables. The dependent variable has a value that depends on something else. For instance, body mass may depend on sex.

Independent variables explain variation in dependent variables and are often experimentally manipulated or assigned. In this case, sex would be the independent variable. Variables can be continuousthat is they can be described, by numbers, with values varying along a continuum. Body mass, measured in grams, is a continuous variable. Variables can be discrete – that is they are described by categories and thus, can be described with a limited range of values. Sex is a discrete variable. Variables can also be ordinal – that is they can be ranked in some way. The results of a race are ordinal; you can’t come in 23.124th, but you can come in 23rd place.

Dichotomous variables are those that take on two mutually-exclusive values. For instance, you can be alive or dead following some treatment. An animal can sing or not sing. It can breed or not breed. (Modified from Dan’s Quick-and-Dirty Guide to Statistics).

Example from the marmot experiment: The independent variable is human disturbance, the dependent variables include the proportion of time spent foraging and/or being vigilant as well as the flight initiation distance (FID). FID is a metric of sensitivity to predation threats. The researcher approaches a marmot and measures the separation distance between human and marmot at which the marmot flees to its burrow.

Example from the mayfly experiment: The presence of brook trout in the stream is the independent variable (also discrete and dichotomous) in the experiment above and the size of the Baetis individuals is the dependent (and continuous) variable.

5. Experimental Design Part II: Design Methods

Once you know what your variables are you can start to design your experimental methods. Your methods and the kinds of analyses you will do constitute your research plan. The research plan defines how you plan to go about answering your questions.

Example from the marmot experiment: To measure the influence of long-term human disturbance on behavior is tricky. Instead of directly manipulating human disturbance, which would be difficult to do in the long term, researchers selected study animals within areas of different levels of human disturbance. This type of study is what is known as a correlational study. Researchers defined human disturbance by counting the number of humans, car and bicycles within a given time period over the course of several days.

Example from the mayfly experiment: Peckarsky et al. wanted to find out whether the chemical cues given off by the presence of brook trout in a stream affected the size of mature Baetis, so they designed an experiment where they exposed some Baetis larvae to stream water without trout (control treatment) and some Baetis to water with trout (experimental treatment). Then they compared the sizes of mature larvae between the two treatments throughout the experimental period.

Important Note! Confounding Variables

While designing experiments, it is important to consider any other variables that may have an effect on your results. Are there other factors – not included in your experiment – that may have an effect on the outcome?

Example from the marmot experiment: Age and sex of marmots can influence behavior. However, in this study experimentally controlling for confounding variables such as age and sex is difficult since the number of study animals in the field is often limited. Restricting data collection to only male or adult marmots, for example, would severely reduce the sample size thereby reducing the ability to detect differences. As done by Li et al., one solution to these issues is to note the ages and sexes of your study subjects and use these data in your statistical analysis.

Example from the mayfly experiment: In order to be certain that brook trout chemical cues were the only factor affecting the Baetis in the experiment, Peckarsky et al. tested other variables such as stream size, substrate size and conductivity to ensure they did not vary systematically between stream of the two treatments.

6. Results

Carefully report your results. Do your results support your hypothesis? What conclusions can you draw based on your results? What questions do you have after the completion of the experiment? (The best experiments ask more questions than they answer. New or unanswered questions form the basis for constructing new hypotheses and designing new experiments!

Example from the marmot experiment: In areas of greater human disturbance, 1) marmots exhibited higher levels of vigilance, decreased foraging and tolerated closer human approaches (i.e. decreased FID) and 2) juveniles tolerated closer human approaches than adults or yearlings.

Example from the mayfly experiment: Peckarsky et al. found that Baetis exposed to brook trout chemical cues matured at significantly smaller sizes than Baetis exposed to water without fish.

7. Conclusions

Always clearly separate your results (the data) from interpretations of the data that form the basis of conclusions. Speculations are also a common inclusion in discussion sections, and also must be clearly distinguished from conclusions supported by data.

Example from the marmot experiment: Despite marmots being more wary in areas of higher disturbance, marmots let humans get closer to them. This somewhat contradictory result may indicate 1) marmots need to acquire more information about predation risk in areas with more humans, and thus increase vigilance, and 2) marmots have learned that humans do not pose a threat and thus tolerate closer approaches. Li et al. also learned that age is an important consideration when examining human impact on animal populations.

Example from the mayfly experiment: The presence of brook trout caused alterations in mayfly size at maturity – this conclusion supports the initial hypothesis.

Having Trouble Getting Started?

Let your creativity be the key. Don’t get too bogged down crafting the perfect question or hypothesis or worrying that you aren’t a “real” scientist. Just ask an answerable question – craft a testable hypothesis – and have fun with your experiment!

digital_bees[1]For inspiration, try reading this peer-reviewed paper highlighting research on bee behavior conducted and written by 8-10 year old children at Blackawton Primary School in Devon, UK.

Next step – Learn more about the effects of fear in a TEDX talk, the Sound of Fear, featuring RMBL researcher Dr. Dan Blumstein.

Attribution

Thanks to undergraduate RMBL researcher Emily Thorne and Nicole Munoz, a PhD Candidate in Blumstein’s lab at UCLA, for major contributions to this section. Thanks also to RMBL researchers Dr. Dan Blumstein and Dr. Bobbi Peckarsky for their reviews and suggestions.