The working hypothesis for this study is that the banana is the funniest fruit. Although the hypothesis is intuitively reasonable and stated as fact in various articles, such as [1], [2] and [3], the hypothesis has not yet undergone serious examination. It's possible that another fruit is actually funnier to more people, or that a less popular fruit is actually funnier but appears in fewer jokes overall due to its lower incidence. This study attempts to measure fruit funniness while accounting for overall popularity or incidence.
Our approach is to use Google to count the number of times a fruit name is found on the Web in conjunction with the word 'joke' and the word 'funny'. For example, 'banana joke' is found 132,000 times, whereas 'banana funny' is found 358,000 times and 'apple joke' is found 445,000 times. This provides a raw hit rate H, combining both popularity and humour value. The next step is to factor out the raw fruit popularity by finding the absolute number of hits for each fruit and dividing that into the number of 'joke' and 'funny' hits. Our final result is the funniness (F) of each fruit, multiplied by 100 for readability.
F(joke) = 100 * H(fruit + 'joke')/H(fruit)
F(funny) = 100 * H(fruit + 'funny')/H(fruit)
Our raw data, as well as F(joke) and F(funny) results calculated from the raw data, are included in the table below.
| Hits (1000s) | Hits + 'joke' | F(joke) | Hits + 'funny' | F(funny) | |
|---|---|---|---|---|---|
| banana | 4420 | 132 | 2.99 | 358 | 8.10 |
| melon | 1330 | 38.8 | 2.92 | 101 | 7.59 |
| strawberry | 3190 | 64.9 | 2.03 | 185 | 5.80 |
| peach | 3620 | 50.3 | 1.39 | 194 | 5.36 |
| cherry | 9300 | 184 | 1.98 | 489 | 5.26 |
| orange | 27700 | 444 | 1.60 | 1400 | 5.05 |
| plum | 2300 | 34.5 | 1.50 | 112 | 4.87 |
| lemon | 7070 | 113 | 1.60 | 318 | 4.50 |
| pineapple | 1610 | 21.3 | 1.32 | 71 | 4.41 |
| grape | 2410 | 38.4 | 1.59 | 102 | 4.23 |
| kiwi | 2200 | 32.9 | 1.50 | 92.4 | 4.20 |
| lime | 3750 | 55.9 | 1.49 | 143 | 3.81 |
| mango | 2200 | 24.1 | 1.10 | 78.5 | 3.57 |
| coconut | 2790 | 34.6 | 1.24 | 94.4 | 3.38 |
| apple | 38100 | 445 | 1.17 | 1250 | 3.28 |
| pear | 2800 | 31.2 | 1.11 | 85.9 | 3.07 |
| grapefruit | 1120 | 12.8 | 1.14 | 32.3 | 2.88 |
Our results confirm that 'banana' is the funniest fruit by either measure, followed by the surprisingly close contender 'melon'. Banana and melon are both much funnier than the third-funniest fruit, strawberry. The three funniest fruits all have a F(funny) greater than 5, and F(joke) greater than 2.
The least funny fruit is either the pear or the grapefruit although both are unfunny by either measure, having very low F(funny) and F(joke). Other unfunny fruits include the apple and the mango (particularly so by the F(joke) measure).
There is a high degree of correlation between the two sets of result. A few fruit were rated significantly less funny by the F(joke) measure than the F(funny) measure, including peach, orange, plum, pineapple and mango. The chart below shows the two result sets charter on normalized axes, to compare correlation.
There are a few flaws with this study. Several such flaws are discussed along with suggestions for future study in this line of research.
In any hit-based popularity study combining multiple words, the result set can easily include hits where the multiple words are only loosely correlated. Proximity searches, or searches for exact phrases, might go some ways to reduce this problem. Human examination of each result is infeasible as the number of hits for each multiple-word search ranges from 12,800 to 445,000.
Another approach would be to count incidence in more closely focused data sources, such as databases of jokes (such as [4] or [5]. However, our quick investigation found that these databases do not have easily parsable hit results. One of them returns hits for 'pear' including hits for 'appear' and 'pearl', which would have to be manually removed from the result set. The other does not count its total number of hits, requiring manual counting.
A harder problem to solve is that of homography. The results for 'apple' and 'orange' are likely to be quite misleading, as a popular computer company with the name 'apple' may appear in jokes, and the colour 'orange' may appear in jokes. Some filtering of the results for computer-related hits is possible, and it may even be possible to analyze the hits grammatically to see if the word 'orange' is used as a noun or adjective, but such analysis is beyond the scope if this study.