Coke vs. Pepsi Taste Test: Experiments and Inference about Cause

This lesson plan and activity are based on material from the NSF-funded AIMS Project (Garfield, delMas and Zieffler, 2007). For more information contact Joan Garfield at jbg@umn.edu

This material was originally developed through CAUSE
as part of its collaboration with the SERC Pedagogic Service.

Summary

The Coke vs. Pepsi Taste Test Challenge has students design and carry out an experiment to determine whether or not students are able to correctly identify two brands of cola in a blind taste test. In the first stage of the activity students design and conduct the experiment. In the second part of the activity students use Sampling SIM software (freely downloadable from http://www.tc.umn.edu/~delma001/stat_tools/) to simulate and gather information on what would be expected under chance conditions (i.e., if students obtained correct answers only by guessing). The students then compare the observed results to the chance results and make an inference about whether a given student can in fact correctly identify Coke and Pepsi in a blind taste test. Finally, the experiment is critiqued in terms of how well it met the standards for a good experiment.


This activity allows students to gain a better understanding of the experimental process and causality through considering control, random assignment, and possible confounding variables. The activity also allows students to begin to understand the process of hypothesis testing by comparing their observed results of the taste test to the results obtained through Sampling SIM (which model would be obtained by chance). Students make an inference about whether particular students in their class can truly tell the difference between Coke and Pepsi by reasoning about how surprising the observed results are compared to the simulated distribution of correct identifications by guessing. The activity also provides an opportunity for discussing generalizability to a population.

Used this activity? Share your experiences and modifications

Learning Goals

  • To learn the characteristics of a well defined experiment.
  • To learn the difference between an experiment and an observational study.
  • To learn to recognize instances of confounding.
  • To learn to understand and recognize instances of experimental control.
  • To learn that randomizing the assignment of treatments protects against confounding and makes cause and effect statements possible.
  • To build the underpinnings of inference.
  • To understand the process of hypothesis testing by comparing the observed results to the results obtained under chance conditions.
  • Context for Use

    This activity can be used in a unit of introductory statistics on producing data through experiments. It could also be used in a unit on one sample tests of proportions.

    Description and Teaching Materials

    This activity takes place in two stages as described below.


    Roles: Within each group of four students, assign each student to one of the following roles.

    • Tasters--those who think they can tell the difference (blind to test).
    • Runners--those who run cups of cola from room to hall (blind to test).
    • Recorders--they record results of tasters decisions about whether they are tasting Coke or Pepsi.
    • Pourers--remain in the classroom as pourers/observers.


    The following materials are needed:

    • 10 Dixie cups per group for taste testing
    • 8 additional Dixie cups for clearing the palate
    • Coke and Pepsi (4 cans of each is enough for 8 groups)
    • Recorder slips for each group
    • Coke/Pepsi pourer slips, where each group is given a random order of Coke and Pepsi over 10 trials
    • Sampling SIM software loaded on the computer or another simulation program or applet (see Available Technologies)</li


    Copy of Student Handout (Microsoft Word 43kB Feb25 07)


    Stage One of Activity: Designing and Conducting the Experiment

    First, students are asked to consider how to design an experiment that will allow them to determine if anyone can correctly identify two different brands of cola in a blind taste test. After a discussion of various methods, a plan is introduced to use in conducting a taste test.


    Students are asked to self identify who can correctly identify Coke or Pepsi in a blind taste test. Groups are then formed with one of these students in each group to be the tester. Groups of four work best. Each group member has one of the following roles:

    First, the tasters, the recorders, and the runners will leave the room. The pourers are produce random series of Coke or Pepsi using a coin to determine what the tasters would taste. The pourers will be the only members of the group knowledgeable of the condition. The pourers will pour the appropriate drinks into paper cups, and leave them in a row to be tasted at their table. These people may switch groups when the taste testing begins so they will not know the order of colas to be tested.


    Next, the runners will bring the first Dixie cups with cola to their group taster. The tasters will taste the drink and make a decision about whether they think it is Coke or Pepsi that they are drinking. The recorder will keep track of the taster's decision. The tasters will cleanse their palettes in between trials by taking a drink of water.


    Repeat the above process 4 more times for a total of 5 trials. At the end. The results are reveled and the number of correctly identified colas is tallied for each taster.


    Stage Two of Activity: Analyzing the class data

    First, students discuss the results, being asked if they think any of the results suggest that a student is doing better than just guessing. They are asked what kind of data would be expected if they were just guessing. This leads to simulating data for the situation of guessing (p = .5, n = 5 trials). Use Sampling SIM (or another simulation program or applet) to simulate the Coke/Pepsi activity, simulating data for 500 trials. Then students can compare the number of correct guesses to this distribution to see if their score is due to chance (in the middle) or surprising (in one of the tails). A final discussion involves critiquing the experiment and talking about what could have made it better (e.g., more tastes).

    Teaching Notes and Tips

    Try to make sure that you have one person in each group that thinks they can distinguish between Coke and Pepsi. Try to have students make conjectures about what they would expect before gathering or simulating data.

    Assessment

    Have students discuss or write answers to the following questions:


  • How were three elements of a good experiment (random assignment, control, and replication) included or not included in this experiment?
  • Describe any possible sources of confounding in the experiment.
  • Can you generalize the findings of this experiment to all students at this university? Why or why not?
  • References and Resources

    NSF-Funded AIMS Project (J. Garfield, R. delMas, and A. Zieffler, University of Minnesota)