While dining on chateaubriand at a swank restaurant in Denver, Marianne Gillette, group leader for the sensory evaluation unit at McCormick & Company Inc., detected something awry in the "homemade" sauce. Hydrolyzed vegetable protein, she told an embarrassed chef. Gillette had just helped develop a brown gravy mix for McCormick and she knew the flavor components of such products "inside out."

After drinking a cup of coffee at a Hot Shoppes restaurant, Elaine Skinner, laboratory manager of the production evaluation department at General Foods, informed the waitress: "It's freshly brewed, high in brazils the type of bean and probably in the neighborhood of a 68-rose colored [the degree to which the bean had been roasted]." Skinner should know. She's been a professional taster for more than 20 years.

There is more to gravy mixes or coffee beans than meets the eye -- or the average nose and mouth, for that matter.

Trained tasters are the human instruments that help major food corporations make formulation changes or new products, analyze the competition or ensure quality control. Called sensory evaluators, they are the fine tuners who bridge the gap between the food science laboratory and the consumer panel.

Particularly at a time when about half of test-market products flop, sensory evaluation -- a science that combines psychology, statistics, experimental design and physiology -- is becoming an increasingly popular tool for companies. David Holmes, president of the contract research facility of the National Food Processors Association, attributes this emphasis to the increased variety of packaged foods available to the public as well as an audience that is more discriminating than ever.

Should McCormick buy more oregano from the Dominican Republic? Does it taste similar enough to the Mexican or Greek variety the company already sells to consumers? What is the shelf life of Cap'n Crunch cereal? At what point in time is a taste difference perceived? Will the hops that Anheuser Busch is using this year produce a brew close enough to the original? These are the types of questions that sensory evaluation units can help companies answer.

Like high school biology students dissecting the body of a frog, the taste panel trainees at McCormick's technical plant in Hunt Valley, Md., are dissecting the components of bacon bits. In-house employes who function daily as secretaries, chemists or accountants, these tasters are also part of an 18-month training class run by the sensory evaluation unit.

As in teaching a child to speak, presenting the language of sensory evaluation simply involves matching up a physical property to a descriptive word (thus, say many sensory evaluators, anybody can learn to be a professional taster, so long as they are interested, can concentrate and have an ability to communicate).

Seated around a table in a room with red lights to mask the true color of the food, tasters are told by group leader Gillette to sample the bacon chips and then go back and smell and taste its "attributes." Its "attributes" or flavor components sit in little plastic cups covered with watch-glass covers.

The tasters sniff and nibble samples of: 1) fried bacon fat, 2) fried pork fat, 3) lean fried pork chop, 4) burnt bacon, 5) properly cooked bacon, and 6) pure maple syrup on a perfumer's stick. Next week, says Gillette, tasters will complete identifying the flavor characteristics of bacon bits by tasting examples of smoke, soy and salt.

Every product has its own set of descriptors, says Jean Caul, who runs a sensory evaluation center at Kansas State University and was one of the developers of the original flavor profile system devised at Arthur D. Little more than 40 years ago. The descriptors, however, may vary depending on the company's own language.

In the cabinets in the sensory lab at McCormick, for instance, are samples, or reference standards, of the "flavor notes" for a variety of spices. Stacked on one shelf are the reference standards for vanilla, which include raisins soaked in rum and a jar of Play-Doh. "Somebody found a Play-Doh note in vanilla," says Gillette. And one of the differences between real vanilla and imitation vanilla is that the imitation often lacks a "beany resinous note," she said.

At Pepsi, Mike Baum, manager of sensory evaluation, said that the flavor notes of that soft drink include "citrus, vanilla, brown spice and pruney." Coke is more "pruney" than Pepsi, says Baum, and Pepsi has more "citrus notes" than Coke.

These words help descriptive analyzers to characterize a product, allowing them to make changes and comparisons. For instance, if Pepsi wants to make a caffeine-free product taste similar to regular Pepsi, the presence and intensity of the flavor notes of the caffeine-free product could be compared by the panelists to the target -- the regular Pepsi. The formulators could then use that information to adapt the formula until the gap between the two has been narrowed.

At the sensory evaluation unit at the National Food Processors Association food laboratory in California (there are about a half dozen private sensory evaluation labs across the country that do contract work), group leader Marianne Lang said that the outfit recently did a project for a pizza company that felt it wasn't doing as well as its competitor. The descriptive analysis, which compared the two pizzas, helped show the company the intensity differences between the cheese flavor, sausage flavor and so on.

Descriptive analysis is only one test method that sensory evaluation units use and one which involves trained tasters. These are the people who can tell the country of origin of the vanilla bean, the season it was harvested and whether there was a fire in the warehouse where it was being stored.

Another method is called "difference" testing. Like foods are more different than people think, says Gillette. "But the real question is, how different is okay?"

For example, if a company wants to replace an ingredient in its formulation with a less expensive one, it will want to know if the difference is perceivable in the final product. Or perhaps a certain flavoring has been banned, and the company wants to know if using the substitute is an acceptable replica.

At Hershey Foods, the company is experimenting with using Malaysian cocoa beans in its chocolate blend (the company uses about 10 different cocoa beans already). Lab technicians cook up a product that partially contains the Malaysian beans. Given samples of the Malaysian-containing chocolate and Hershey's originial blend, the difference panelists are asked to identify which is which, according to the company's manager of sensory research, John Riti.

Another type of testing is called "acceptance" or "preference." The results of this test will tell a company just that: which product is preferred. Often this is done between competing products.

The same day McCormick was training tasters how to taste bacon bits (the company usually has about four or five tastings on its calender per day), it asked other employes to preference taste two competing tortilla chips.

The panelists enter a room that is sectioned off by individual taste booths that resemble library carrels. After lifting up a rolltop contraption in front of them, the tasters find two tortilla chip samples on a tray in little plastic cups labeled KE and YL. Also on the tray are a pencil, a cup of mineral water to clear the palate between chips, a candy treat for performing the test and a form that asks the taster to check the sample preferred in terms of color, texture and overall flavor.

On the other side of the wall is the sensory lab, where technicians are busily lifting up rolltops on their side and placing samples on the trays. The labeled cups are alternated between trays so that the tasters don't taste the samples in the same order. Neither side sees what the other is doing; the only contact beyond the common wall is the sound of crunching chips.

Using preference tests or descriptive analysis to see what the competition is doing is a common practice that companies are hesitant to discuss specifically. (Not to mention their flavor clients or reformulation plans.)

Gillette, who says that many companies are trying to figure out the formula in Heinz ketchup and Kraft French Dressing, says that McCormick looks at "hundreds" of competitors' products every month. And Klaus Zastrow, vice president of brewing technical services at Anheuser Busch, a company that has two taste panels at each of its 11 breweries, plus a handful of corporate taste panels, said the panels will periodically conduct a blind tasting to see "how does Miller or Stroh's or Coors taste these days?"

Taste panels also help companies monitor their own products. Zastrow said that tasters "buy back" Anheuser Busch beers from retail outlets to see if they have remained fresh through the distribution system. Not only that, but the company will take beer samples and "punish" them, subjecting the product to a variety of temperature changes to see how and when the taste is effected.

This quality control, or shelf life component is a common use of taste panels. Pepsi did it with aspartame, the sweetener that loses its punch after extended, unrefrigerated shelf life. The company wanted to know at what point in the product's life can people detect a decline in sweetness. The panel's findings were then used to gauge how fast the product must move to the consumer and how much bottlers should manfacture at once.

And Quaker Oats uses its taste panelists to monitor flavor changes due to temperature variations and packaging materials. According to June Yantis, senior manager of the sensory evaluation group at the company, Quaker has subjected Cap'n Crunch cereal to "humidity cabinets" meant to simulate Florida.

But companies aren't always measuring flavor differences. At Quaker, for example, taste panels may be measuring crispness of a cereal. If the company has a crunchy cereal as its target, it may test it to see if it stands up against milk during different periods of time.

And tasters may be asked to be sleuths, looking for "off notes" in a product, too. Lang at the NFPA facility says the lab is presently looking at a nonfat milk that may have a "goaty" character to it. And according to Holmes at the same lab, a jelly bean company recently asked that an odd taste in its product that had forced a shut down in production be identified. The descriptive panelists identified a "machine oil" quality to the jelly beans, and using that tip, the plant pinpointed a leaking oil seal in their operation.

In addition, Lang says the lab is frequently called upon to taste test for insurance claim disputes. A chocolate company, for instance, claimed that its product had picked up "cherry notes" while being shipped next to a load of cherry flavoring. And then there are the cases of damage to a warehouse by fire or flood. The lab's mission: determine whether the sensory characteristics of the food have been lost.

Or the company may be test tasting to make a particular product claim. At McCormick, for instance, tasters are presently tasting freeze dried spices to see if and what claims can be made in comparison to fresh.

While companies vary the testing method depending on their desired outcome, taste panels test the waters before the product may go to a consumer panel, a much costlier testing procedure. Still, the tasting panels are only one step in the making of a product. Marketing, advertising, promotion and the niche that a product fulfills are all key factors.

And after all that, a product may still fail. According to Gillette, a McCormick product marketed as Skillet Magic is now remembered at the firm as "Skillet Tragic."