The episode of high food prices in the period from 2007 and the more recent post‐lockdown commodity price resurgence have prompted discussion of commodity price cycles. Tropical commodities, particularly tree crop commodities, stand out as the most likely to exhibit cyclical patterns. The paper uses annual data for cocoa, coffee and sugar extending back into the middle of the nineteenth century to attempt to identify long cycles. I compare results obtained from the bandpass filter decomposition with those generated by the unobserved components model. Both procedure rely on separation of trend and cycle. This is straightforward in sugar, where the trend is well defined, but more problematic in coffee and, especially so in cocoa. Some evidence is found for a 25 year cycle in all three commodities.