i went down to new york last week for the french culinary institute's harold mcgee series, which consists of mcgee talking about the science of cooking and nils noren and dave arnold from fci showing off a select portion of the fancy stuff they've been doing in their copious spare time to illustrate (and push beyond) the known science of cooking.here's nastassia lopez's official account of day 1 of the series on cooking issues, the FCI's R&D blog. i'm still processing the three day show and tell of the output from one of the cutting edges of food, but these are some early thoughts about the uses of experimentation in cooking.
we got to try some of almost everything that was talked about so that we could compare, for example, the flavour and texture of mashed potatoes cooked with a pretreatment to set the starch granules (cooked to 160F, cooled to 40F, then boiled and passed through a tamis) and mashed potatoes cooked in the traditional method (ie, just boiled and passed through a tamis). [i liked the conventional cooking method better]. there were lots of experiments and it was sweet to be able to try several different ways of preparing the same item side by side. so that was great.
throughout, nils, dave, and harold repeatedly noted that the techniques were not being presented as the one best way to do any given thing (cook a steak, burger, or whatever), but rather as ways to break down the process of cooking as much as possible so that the cook can, given a clear idea of the final outcome desired, achieve that outcome with the greatest possible precision and reliability. compared to the whizbang demonstrations, this relatively theoretical consideration didn't seem top of mind for most of the participants. (in fairness, it's difficult to focus on things like that when being constantly plied with really tasty stuff--see below).
tasty stuff: i think this was a case full of winesap apples, juiced with a pectinase, clarified with a centrifuge, and then evaporated under a vacuum. the bottle is about the size of a roll of quarters and contains an apple syrup that tastes like getting hit in the head by a case of apples.
more tasty stuff: egg on egg on egg (ikura on egg whites, served on a round of all-egg bread cooked in a pressure cooker).
the uses of experimentation
this was the last thing on my mind when i showed up on the first day, but it turns out you can't understand how experimentation is valuable to cooks without thinking through the assumptions made when experimenting in food.
the series is one of the most basic instances of the experimental method in cooking: all variables but one are held constant to illustrate the effect of that one variable on the outcome. the example from thursday morning's discussion of heat in cooking was a series of pieces of salmon cooked at different temperatures, to illustrate how temperature has a nonlinear effect on the texture of cooked salmon. based on the series, cooking salmon to between 48C and 52C produces a safe and tasty piece of salmon. but this is not the most useful conclusion to draw from the experiment. the real takeaway is understanding that cooked salmon has five major discrete textures, and that these textures change over relatively small temperature intervals.
the salmon temperature series.
the salmon temperature series described above takes a complex real system (cooking potentially different types, cuts, sizes of salmon pieces at varying temperatures, pressures, for varying amounts of time, to say nothing of flavourings and cooking media) and reduces it to a simple model that can be experimentally manipulated (the same piece of salmon [sort of] + different temperatures = a range of cooked salmon). the implicit assumption is that temperature is the variable with the biggest influence on the texture of cooked salmon in the real world, and that the relationship between these variables in the experimental model closely approximates the real world.
all experiments that take the form of simplified models rely on at least these two assumptions: 1) that the experimental model has the important variables in it, 2) that the relationship between the variables in the model is relatively close to the relationship between variables in the real world. if those conditions hold, experiments based on simplified models are useful in practice not just because they provide precise and specific guidelines for cooking (even though that may sometimes be true), but also because they sensitise cooks to the materials and tools they work with. part of this is using the experiment's results to understand when it makes sense to pay attention and what it makes sense to pay attention to.