As we just discussed, it's possible for all the methods of a task to fail, in which case the call itself fails.  In many cases, that's a bug, so Step normally treats that as an error.  If we try running [Dog morris] below, we get an error:

# Try: [Dog morris]
Dog benji.
Dog lassie.

Cat morris.
Cat garfield.

However, as we've also just seen, it's useful to allow calls to fail and use the success or failure of the call as a way of answering questions.  Tasks that exist for the purpose of answering questions are called predicates.¹ 

The [predicate] declaration

Tagging a task with the declaration [predicate] tells Step not to consider a failed call to that task to be an error.  That's all it does.  So in the example:

# Try: [Dog morris]
[predicate]
Dog benji.
Dog lassie.

[predicate]
Cat morris.
Cat garfield.

Calling [Dog morris] will fail, but it won't be treated as an error that stops the program, because Dog is declared [predicate].  Instead, the system just responds “no”: Morris is not a dog, given what the system knows.

To make a task be a predicate, you only have to include the [predicate] declaration before one of its methods. That will apply it to the task more generally.

Predicates don't print

The primary use of predicates is to answer questions.  So they generally aren't written to generate output.  There's nothing preventing you from writing tasks that print but are labeled [predicate], but you usually wouldn't want your question-answering tasks to print things.

Notes