Coke/Pepsi Challenge Results

New Years Eve held the Coke/Pepsi 3-shot challenge pitting the discerning palettes of my guests against a gauntlet of samples.  The preponderance of guests stated they’d be able to tell them apart at a statistically significant level.

Purpose: Determine if a tester can accurately identify samples of Coke or Pepsi when presented in lots of three.  Success will be defined as properly identifying all samples in six of nine lots of three.


  1. A pseudorandom sequence of ones and zeros were generated as the seed data.  Zeros were Coke samples and ones were Pepsi samples.
  2. Cups were numbered sequentially and filled with approximately 1-2 oz of the appropriate beverage as listed.
  3. Testers were allowed to take calibration samples of each beverage from a control bottle.
  4. Each tester received 3 sequential cups and after sampling, received a data sheet requesting the cup number and the corresponding observation.
  5. To be statistically significant, each tester had to consume 9 sets, but after much bitching and complaining, we stopped at 7.

Results: After 108 samples tested, approximately 50 samples were properly identified breaking the Colobus Barrier.  Of the 36 lots tested, only 3 went 3 for 3 correctly identifying samples, once again doing worse than chance.  No tester was able to identify samples at a level above chance.

Considerations/Stuff People Bitched About: The soda was flat; each specimen was poured from a bottle that had been open within the previous 30 minutes.  The soda was warm; this was done specifically at the request of Pat Toye and may have partially caused the previous accusation of flatness.

Interlab Repeatability: If anyone would like to stage a repeat under similar conditions with a changed variable of temperature, brands, specimen (cup) size, participants or tester, please contact me.