Enigmatic Bills of Fare (EBoFs) frequently share the same lines, suggesting that there is a transfer of riddles from supper to supper. This raises the immediate question of how and why this is happening. Are riddles being shared based on their physical location? Are riddles being shared in a timely fashion, such that recurring riddles tend to come up in bursts and then fade away? Or is it random: Do riddles just pop up randomly as a sometime fad?

To represent the transfer of riddles from EBoF to EBoF, we constructed a network representing our data set of EBoFs and the lines (individual riddles) shared between them. This data set includes 34 EBoFs (composed of a set of lines) and 246 lines that appear in at least two EBoFs.

For example: The Divine Part of Man appears in 23 EBoFs, including BOF_13. To see how this line fits in, here is an excerpt from BOF_13:


  1. The ornamental part of the head.
  2. The divine part of man.
  3. A tailor’s implement.
  4. A blockhead.
  5. The Grand Seigniors Dominions.


First, of course, can you figure out answer to The Divine Part of Man? In other words, the meal that is represented by the riddle?

It’s heart, or sole.

At any rate, the EBoF goes on; there are 39 unique lines contained in this riddle alone. Each line has its own answer, representing a dish to be served at that enigmatical dinner. Now, on to the graph representation of these connections:

Social network representing shared lines of ebofs

Each node in this network represents an EBoF, labelled with its unique ID. (You can see BOF_13 in the top center of the graphic.) The edges between nodes represent a shared line between two EBoFs, and the weight of the edge signifies the number of lines shared between two EBoFs.

Once the network was constructed, we used the Louvain community detection algorithm to determine clusters and the modularity score of the network. Each node is colored by the cluster it is a part of (there are four unique clusters). The green cluster contains 13 nodes, the purple cluster contains 16 nodes, the orange cluster contains 5 nodes, and the blue cluster only contains BOF_3, which had no shared lines with any other BOF.

Clusters are divided very cleanly by period: 75% of BOFs in the purple cluster are post-1824 (the median year of publication), 92% of BOFS in the green cluster are pre-1824, and 80% of EBoFs in the orange cluster are post-1824. This is a strong indicator that proximity by time is an influence on how lines from BOFs are distributed. In other words, it is likely that riddles from EBoFs were extremely temporal, and maybe even “trendy.” The best riddles likely moved from dinner to dinner in a short span of time before fading into obscurity.

Tested purity by publishing location was less revealing because most EBoFs from this set were published in England. Every cluster had roughly 60% purity by region, with all being majority EBoFs from London.

Find out more here. To stay involved, follow us on Instagram! You can also reach out to @McGillLib, @McGillRoaar or email theriddleprojectmcgill@gmail.com.