There’s something magical about late spring strawberries and fresh summertime corn: flavor! It’s no secret that cooking with fresh ingredients brings great flavors to foods. Limiting your shopping list to in-season items is a fun way to challenge yourself, and because in-season ingredients tend to pack more of a flavor punch, it’s easier to make your cooking taste amazing. There’s another perk of using in-season ingredients: they’re generally priced lower, based on the laws of supply and demand. Grocery stores have to figure out how to sell all those zucchinis when they come up for harvest!
Next time you’re at the grocery store, take note of what new fruits and vegetables have arrived and what is in dwindling supply. Corn on the cob is one of the most seasonal items where I live, nearly impossible to get out of season. Other produce, like peaches, is available in my local store almost year-round, but rarely delights and usually disappoints. Try this cooking challenge for inspiration: treat any food that’s outside its growing season as off-limits. Peach pie in April? Out. Even if you can get a peach in April, it won’t have the same flavor as a mid-summer peach, so your pie will invariably taste bland.
Of course, not every item is a seasonal one. Cellar onions, storage apples, and pantry goods such as rice, flour, and beans are year-round staples. If it’s the dead of winter and there’s a foot of snow on the ground (incidentally, not the best time to eat out at restaurants specializing in local, organic fare), finding fresh produce with good flavor can be a real challenge. There’s a reason winter meals in cold climates lean heavily on cooking techniques to produce flavors. Classic French winter dishes like cassoulet (traditionally made with beans and slow-cooked meats) and coq au vin (stewed chicken in wine) use cellar vegetables and meats from domesticated animals. But come summertime? A quickly sautéed fish with fresh greens is amazing. I can’t imagine eating a heavy, rich cassoulet in the middle of summer, yet in the dead of winter, nothing’s better.
We’re lucky to live in a time with an amazing food supply. Many cuisines are defined by seasonality and the history of the associated region’s food environment. The 19th century French favored dishes like cassoulet and coq au vin based on their food supply. Cuisines in costal parts of the Scandinavian region were constrained by the lack of a road system until only a few decades ago, so it’s no surprise that modern Nordic cuisine incorporates simple cheeses and preservation methods like cured fish while shying away from complex spices.
On the downside, our modern food supply means we’re no longer constrained by seasonal ingredients, which makes it harder to learn how to cook well. Shopping at a farmers’ market can be a great source for seasonal inspiration and flavorful ingredients that will inspire. Consider the seasonal soups on pages 116–118. Buying butternut squash in July is almost impossible, and I wouldn’t make gazpacho in winter. Same thing for seasonal salads. A summer salad with mozzarella, tomatoes, and basil (see page 114)? Yum. A wintertime salad with fennel? A fall harvest salad with toasted pumpkin seeds and sprouted seeds? (Guess who’s hungry now, as I write this!) Understanding flavors from the perspective of the seasons can be as easy as strolling through the produce aisle and conjuring the inspiration by exploration covered in the prior section, if you keep your eyes open to the possibilities.
“Below legally allowed levels” doesn’t mean “100% guaranteed,” regardless of whether you’re buying organic or conventional. In the US, the FDA inspects less than 1% of imports (as of 2012), and excess pesticide residues have been found in some foods imported from abroad when tested by independent researchers. Enforcement (and funding for it) needs to be stepped up.
Computer, what goes well with Tea, Earl Grey, Hot?
Technology isn’t that far away from being able to answer questions like this, and it’s exciting! Imagine opening your fridge door, seeing a few leftover ingredients from a prior night’s meal, and tapping a button on a device and saying, “Show recipes that use chicken, cilantro, and lemons.” My current high-tech watch already lets me do this!
What about more inspirational possibilities? Finding recipes that use a handful of ingredients is one thing, but what if we could computationally predict new recipes, creating combinations that had never been tried before yet delight with flavor? Thanks to a better understanding of how flavors work and having enough combined data to comb through, it’s now possible.
There are two main approaches to computational flavor inspiration: co-occurrence of ingredients and chemical similarity between ingredients. Both have their advantages and disadvantages; more recent research that combines the methods is beginning to bear fruit. Computers are really good at comparing lots of numbers, and these methods definitely benefit from that. (They’re also really good at doing exactly what they’re programmed to do, not what we necessarily meant them to do.)
First, a disclaimer: picking pleasing flavors—or at least ones that invoke an emotional response or trigger a memory—is somewhere between an art and a science. No scientific equation can capture the entire picture, and what you’re craving at any given moment will also vary. Still, understanding how “flavor compatibility algorithms” work can provide you with a way of organizing your thoughts on food, and the results can be useful for the more inquisitive, off-recipe type of cook.
If items A, B, and C go together in one dish; and another dish uses B, C, and D; then there’s a decent chance that A might work in the second dish as well. These sorts of transitive relationships aren’t guaranteed to work, but they are useful enough that most good cooks use them intuitively. Say you like guacamole and know that it commonly contains avocado, garlic, onion, lime juice, and cilantro. When tossing together a salad that has similar ingredients—say, tomato, avocado slices, and onion—it’s reasonable to guess that some coarsely chopped cilantro will work well in it, and maybe even some crushed garlic in a vinegar/oil dressing.
What if we took this idea a step further by computationally examining thousands of recipes and their ingredients? A few projects and notable books have already done this, but it’s still a fun exercise. Snag a few thousand recipes (easy enough for a computer science major like myself), run them through something to clean up the data, and voilà! You’ll have a co-occurrence matrix that shows the relative probability that an ingredient shows up with another. With a few tweaks (normalizing the weights to be 0 to 1; dropping salt, which links everything together), the results become almost human readable. (See http://cookingforgeeks.com/book/cooccurrence/ for a .csv file.)
For chocolate, the most common ingredient it’s paired with is vanilla (giving it a weight of 1 in the normalized co-occurrence matrix). The second most common ingredient (in my dataset) is milk (0.320). Walnuts (0.243), oil (0.166), cream (0.128), and pecans (0.121) are the next four. Hearing this list yields no surprises; chocolate is common with vanilla, dairy, and nuts. For other ingredients, it can be a boon. What spices are typically used with beef? (Black pepper, parsley, thyme, bay leaf, oregano, chili.) Or chicken? (Parsley, thyme, basil, paprika, cayenne pepper, ginger.) You could extend this by collecting different collections of recipes—how does ingredient co-occurrence change between cultures? (“Computer, adjust recipe to be Tex-Mex style.”) Or over time? Imagine the possibilities.
Data is only as useful as the ability to see and act upon it, so unless you’re a spreadsheet junkie, something more is needed to visualize the data. I’ve hacked together a simple interface for clicking through the various ingredients (see http://cookingforgeeks.com/book/foodgraph/). With time, it’ll become unusable (or go offline). There’s an expression in software, “code rot,” that describes how software becomes more buggy and less usable as the systems we use are updated and no longer 100% backward compatible. See the example image for chocolate to understand what the software looks like if you’re unable to load it.
Many of the compounds that give ingredients their flavors can be measured and quantified—assuming you have access to lab equipment that does stuff like chromatography! Drop a sample in, separate out the compounds, and compare the results to those of known compounds. Okay, okay; this is a gross oversimplification. Maybe some day in the distant future my watch could do it, but for now it’s not easy; and even the best lab equipment is not sensitive enough to detect all the odorants our noses can smell. For this reason, I’m going to describe the concept of chemical similarity using odor descriptions instead; we’ll use our noses as the chemical detectors. It’s only one step removed from measuring the odorants directly.
One way to determine similarity is by measuring a number of different variables—say, quantities of potential compounds or odors—and then comparing items based on those different variables. It’s a two-step process: first, figure out a bunch of numbers that describe an individual item; and second, compare those numbers between different items.
This is more easily described with an example. Imagine a flavor profile for a food item, where the profile is how much the food item smells like the terms in Andrew Dravnieks’s 146-odor list (see page 94). For every term in the list, take an item of food and score it on a scale from 1 to 5, where a score of 1 indicates “doesn’t smell like it at all” and 5 is “the very definition of the word!” Given a pear, how much does it smell like a “heavy” odor? 1. Fruity? Maybe a 3? Or how about fragrant? Say it’s a ripe pear, so 4. (The full odor atlas that Dravnieks created does something similar using a collection of known chemical compounds.) This first ranking step is not asking if the food item and odor description are compatible, just if the odor label accurately describes the smell and quantifying it with a number.
The second step to similarity matching compares the values for different ingredients, based on the theory that ingredients with similar scores can be combined or substituted for each other. Given the scores for the odors you sense in a pear, and those you sense in a banana, how much overlap would there be? You can plot a graph (almost like a histogram) for the two, showing how similar they smell. Do this for a bunch of ingredients, and it’s easy to show that pear and banana have more similar odors than, say, salmon and pear.
Unlike ingredient co-occurrence, the chemical similarity method can find overlapping flavors that wouldn’t historically exist. You can imagine a graph with all the ingredients for a dish, showing all the “frequencies” present in the smells of each ingredient. Think of it like the various instruments that contribute to a piece of music: each has its own set of frequencies, and the combination of all the instruments makes up the overall song’s frequency distribution. When in tune, the frequencies line up: different ingredients hit the same odor terms, and not too many odor terms are struck. And as in music, when a dish is out of tune, the combination will be jarringly dissonant, even if each item is fine individually.
Of course, this music analogy isn’t a perfect fit for thinking about flavors: chemical changes brought about by cooking or by reactions between compounds in the foods change the histogram. The music analogy also doesn’t cover other variables in foods, such as texture, weight, or mouthfeel. This method works best with ingredients whose primary purpose is conveying odor. Soups, ice creams, even soufflés: all are methods of transporting the flavors and aromas of ingredients without carrying the texture or volume of the original ingredient.
Heston Blumenthal, the UK-based chef best known for his restaurant The Fat Duck, has used a number of novel flavor combinations: strawberry and coriander, snails and beetroot, chocolate and pink peppercorn, carrot and violet, pineapple and certain types of blue cheese, and banana and parsley. They sound crazy, but the research supports them and they’ve worked in his dishes.
Many chefs—often pros, but also non-pros who’ve been cooking for years—can imagine flavor combinations in their heads, doing something similar to this process mentally. Just as a composer imagines each voice and track in a piece of music, an experienced cook imagines the profile of the entire dish, from the appearance to texture and aromas. Good cooks think about which notes are missing or are too soft and figure out what ingredients can be added to bring up those values or bring down others.
What about achieving entirely new pairings, combinations that have no precedence in tradition? That’s where this method shines. Research chefs searching for new ideas spend an inordinate amount of time working on new flavor combinations. Some top-tier restaurants run research kitchens, devoted to laboratory work and staffed by individuals holding both master’s-level degrees in hard sciences like chemistry and degrees from top-tier culinary institutions. For novel high-end restaurants and the packaged food industry, coming up with new flavors can be extremely lucrative. While the more unusual combinations they come up with may sound unappealing or call for uncommon ingredients—how often do you have caviar on hand?— they do work. At the very least, you might find these types of tools a fun source of inspiration to try new things. Experiment!
GRAPH USED BY PERMISSION OF BERNARD LAHOUSSE OF FOODPAIRING.COM