I have a question about why my cheese sauce periodically curdles. I have a recipe for a basic Mornay sauce -- make a roux with equal parts butter and flour, cook until golden, whisk in milk, heat until just below boiling, slowly stir in the chopped cheese.

Sometimes it works fine; other times, the cheese never melts, but rather the sauce separates, creating lumpy cheese globs or strings and a wheyish liquid. Any thoughts on what could be causing my inconsistent sauce making?

Many cookbooks will tell you that you have allowed the sauce to become too hot before adding the cheese, and that's true. But they never tell you why the sauce curdles. That's what I'm here for.

Cooking with cheese has always been a challenge, as might be expected from the three major constituents that make up about 90 percent of most cheeses: fat, protein and water -- substances that resist being combined into a homogeneous mixture.

These three constituents, along with some calcium and other minerals, come from the milk from which the cheese was made, of course. But their proportions have been drastically changed. Milk is about 88 percent water and only 3 percent each of fat and protein. A typical Cheddar cheese, on the other hand, consists of 37 percent water, 33 percent fat and 25 percent protein. This change in composition is accomplished by coagulating the milk's protein with lactic acid or an enzyme, whereupon the resulting curd drags the fat along with it, leaving most of the milk's water behind as whey.

So much for the briefest course in cheese making ever devised.

Real Cheese

Over hundreds of years, humans have figured out hundreds of ways to blend those three major ingredients into homogeneous solids or semi-solids, ranging from soft (Brie, Camembert, Neufchatel) and semi-soft (Muenster, Port du Salut, Roquefort) to hard (Cheddar, Swiss, Gruyere) and very hard (Parmesan, Romano). But as soon as we try to melt a hard cheese, we have to provide just the right conditions to keep its protein, fat and water blended together without separating into a mass of curdled protein soaking in a puddle of insipid whey.

If we are to succeed in that enterprise, we must liquefy the fat in the cheese, while at the same time forcing it to mix with whatever watery liquids there may be, whether wine (as in cheese fondue) or, in your case, starch-thickened milk.

Heating softens the cheese's fat until it gradually becomes a thick liquid. The problem is that at the same time, the heat is coagulating the proteins. Their twisted, ribbon-like molecules become unraveled by the heat-induced agitation, whereupon they are free to bond with one another (cross-link), forming a complex, tangled network that shrinks tighter and tighter as the heating continues. If the heating is carried on long enough or to a high-enough temperature, the tightening protein network will ooze fat and water like a squeezed, wet sponge. Result: tough lumps of protein swimming in an oil slick worthy of the Exxon Valdez. And when these lumps are stirred, they tend to stretch out into elastic strings.

The trick, then, is to melt the fat without coagulating the protein. The cheese must be melted at as low a temperature as possible, ideally no higher than 160 to 170 degrees (even a double boiler is risky), and held at that temperature for as little time as possible. In your Mornay recipe, you must allow the thickened milk to cool to those temperatures before adding the cheese. Heating to "just below boiling" was your first mistake.

To minimize the heating time, it is essential to shred or grate the cheese before adding it, gradually, to the other ingredients. Tiny bits of cheese have a huge amount of surface area exposed to the hot sauce, so they melt faster and at a lower overall temperature than larger pieces. "Chopping" the cheese isn't enough. That was your second mistake.

Now what if, because of a few moments of diverted attention, say, to answer the telephone, you return to face disaster: a stringy, watery sauce? No guarantees, but there may be a way out. A technique attributed to the late culinary dean of American cookery James Beard is to briskly whisk in a little lemon juice. It can be expected to work because protein molecules need low acidity and calcium ions -- positively charged calcium atoms -- in order to aggregate into their tangled networks. The lemon juice's citric acid interferes with this aggregation in two ways: It increases the acidity and it ties up (chelates) the calcium ions.

One measure of Beard's intuition is that he probably never even knew what a calcium ion was.

Processed Cheese

We are blessed (?) to have many other options for adding cheese, a much-loved flavor brightener, to our foods. Dozens of cheesy (no offense) products beckon from our supermarket shelves. I will list their FDA-defined categories here in order of diminishing faithfulness to the historic and revered concept of cheese. Besides flavor, their primary virtue is that they are all easily meltable and blendable, because they usually contain emulsifying agents and will have been beaten into submission long before they reach your kitchen.

Pasteurized process cheese: A mixture of two or more cheese varieties that have been heated and blended together with an emulsifier and optional ingredients such as water, salt or coloring with a minimum of 47 percent milk fat. These cheese products may contain added cream or fat, making them more meltable, but they must be at least 51 percent actual cheese. Example: most American cheese.

Pasteurized process cheese food (note: not a "cheese" but a "food"): These products contain enough added ingredients such as cream, milk, skim milk, buttermilk or whey to reduce the percentage of actual cheese in the product. They may contain emulsifiers such as phosphates, citrates or tartrates, but must contain at least 23 percent milk fat. Example: Land O' Lakes Individually Wrapped Singles.

Pasteurized process cheese spread: A pasteurized process cheese food that may contain a sweetener plus stabilizing and thickening gums such as carrageenan. Must contain at least 20 percent milk fat. Example: Kraft Olive & Pimento Spread.

Pasteurized process cheese product: Any process cheese product that doesn't meet the minimum fat contents above. Examples: Kraft Singles, Velveeta.

Imitation cheese: Made from vegetable oil. Minimum milk fat: 0 percent. In a glass by itself is Cheez Whiz Cheese Sauce. After whey, its most abundant ingredient is canola oil. Milk fat? Less than 2 percent.

Orange glop: Not an official FDA classification, but the name I give to the stuff poured over nachos, French fries and hot dogs in places where I wouldn't eat.

And consumers are supposed to think they're all simply "cheese"?

Robert L. Wolke (www.professorscience.com) is professor emeritus of chemistry at the University of Pittsburgh and the author, most recently, of "What Einstein Told His Cook: Kitchen Science Explained"

(W. W. Norton). He can be reached at wolke@pitt.edu.