Why a spreadsheet? Why not a Word doc or a domain-specific language or a graphical model?

The original Category-Partition Method used a specialized notation, what today we'd call a "domain-specific language." In that way, it was very similar to today's Gherkin (the language used for Cucumber).

However, text files and DSLs are easy for some people, but they can also be difficult to learn and to use. I found that when matching inputs and expected results

So what happened to Executable UML?

It's in there! Look closely at a multi-step (multi-tab) test spec and you'll see that it's really a state machine. The tabs are the states, the actions are the events, and the results that connect actions to other tabs are the transitions. Check out our microwave oven demo to see how an Executable UML state model can be represented as a TAME spec.

What about the information model (class diagram)?

Although TAME is most often used for testing functional behavior ("do this, then do that") it can also be used for forming combinations of data values. Here's an example: Suppose you have a class with three attributes. Each attribute can have several different kinds of values. Different values, or combinations of values, need to be tested as they will cause different results.

Make a test spec for the class. Each attribute is a distinct input and the different kinds of values are the choices for the input. Then define the different outcomes or situations as distinct result columns. TAME can now generate different combinations of values.

And relationships?

They can be done as inputs as well. See our forthcoming whitepaper.