Wednesday, March 17, 2010

The Apple Test

Thanks to @iszi_lawrence and @SkepticBarista for pointing me towards this bit of nonsense. You cut an apple in half, bung one in a bottle marked "love" and the other in a bottle marked "hate". You then love the one half however you see fit, and hate the other. As the Daily Mail's photo dramatically shows, the hate half will rot faster than the love half. Let's see how we might put it to the test properly.

We need to run the experiment on more than one apple. If we only do this once, it's not going to tell us very much. We're expecting a binary result, that each apple will have a marked difference in quality by the end of the week. We'll only get meaningful data, then, through repeated experimentation.

Blinding - we're going to use bottles that are identical, but each has a collar, with "love" or "hate" attached. The tops of the bottles will feature a randomly assigned number. The randomly assigned numbers are in pairs that are already randomly assigned as either "love" or "hate". This will become very important later.

We fill the bottles with water. As long as it's the same kind of water in both bottles, whatever we use ought not make much impact on the test. We then cut the apple in half. It may be worthwhile examining ways in which we can ensure that an identical procedure is used for this, so that comparisons can be made across different apples. However, for the purposes of comparing the "love" and "hate" halves against each other, as long as the apple is cut more or less into two equal sections, it oughtn't affect the test due to the randomisation. We would expect any biases in cutting to be spread out.

The apple halves are placed into the bottles and sealed. It is at this stage that we discover which bottle is to be "love" and which "hate", by checking the number on the bottle cap. This process can be kept secure by giving this job over to a particular individual, rather than allowing each "apple host" to have access to the codes for their bottles. We only establish the love/hate bottles after the apples have been sealed into the bottles because we do not want to introduce a possible bias. We don't want experimenters to decide based on any kind of aesthetic reasons, or through the early browning of the apple half prior to it being introduced into the bottle, as this will skew the results - we will no longer be comparing like with like.

The collars are placed on the correct bottles. Next we need to isolate the thing that we are looking to examine, which is the application of "love" and "hate" to the apple halves. This means that the only thing different between the treatment of both halves will need to be the emotional attachment created by the experimenter. Therefore, the halves need to be kept together, or in separate containers that are kept at the same level of light, heat, humidity, noise, etc.

You may wish to formalise the way in which the love and hate bond is created. This could be through, perhaps, addressing the love jar with kind words and compliments for half an hour a day, before turning to the hate jar and insulting it for half an hour a day. Keep in mind the requirement for maintaining like atmospheres for both jars. There must not be handling of the jars, unless one can guarantee that the jars are handled in an identical manner.

Once the week is up, the collars are removed, and each pair is judged in turn. The judges, who will not be aware of the love/hate assignation, must decide based on examination of the apple halves, which half is in a better condition. It is vital at this point that any possibility of feedback is removed. Ideally the apple hosts should remove the collars before passing both bottles together to someone who in turn passes them on to the judges; this reduces the possibility of any cuing, subconscious or otherwise, that will give the judge a better idea of which is the love half, and which the hate.

Only after they have made their judgements can we compare their results with the actual love/hate halves.

By chance alone, when this experiment is run in sufficient number, we would expect a fairly even spread - we would expect roughly half of the "love" apples to be in the better condition. If we find that there is a bias in distribution of conditions, then we need to examine our protocols to ensure that "leakage" is not taking place anywhere; i.e. that the randomisation process and the blinding process are working as they should. If we are certain that our experiment is holding true and giving us reliable data, then we publish our findings, and invite our peers to decide if the results of the experiment are meaningful. It is not unusual that our peers may at this stage uncover flaws in our methodology that have led to the results. Where this is the case, we look at how we may eliminate those flaws in further experiments.

Given sufficient funding, I will be happy to undertake this experiment and shed some light on whether or not Nikki Owen is on to something.

1 comment:

  1. Thanks for noticing this...course if hate was really that powerful, the average Daily Mail reporter would be unable to enjoy apples at work, as they'd rot away to nothing as soon as they entered the office.


Please keep comments on topic, and be respectful of one another.