Now we can start to define other relationships:

# Try: [Grandparent stan ?who]
[predicate]
Grandparent ?c ?g: [Parent ?c ?p] [Parent ?p ?g]

[predicate]
Parent bart marge.
Parent bart homer.
Parent lisa homer.
Parent lisa marge.
Parent stan sharon.
Parent stan randy.
Parent randy grandpa.
Parent jimbo grandpa.

Which says that ?g is a grandparent of grandchild ?c if ?p is a parent of ?c, and ?g is a parent of ?p.  Note that we’ve adopted the convention that the grandparent is the second parameter.  But we could have chosen it to be the other way around; the computer doesn’t care.  The code for Grandparent looks almost the same as the code for Sibling — it's just a change in how the parameters match up between the calls — but it gives us a different choice tree and solutions:

graph TB
    start("[Grandparent stan ?who]") -- "<b>Grandparent ?c ?g: [Parent ?c ?p] [Parent ?p ?g]</b> <br> ?c = stan <br> ?g = ?who" --> p1("[Parent stan ?p]")
    p1 -- "<b>Parent stan sharon.</b> <br> ?p = sharon" --> p1sharon("[Parent sharon ?g]") --> f(Fail)
    style f fill:red
    p1 -- "<b>Parent stan randy.</b> <br> ?p = randy" --> p1randy("[Parent randy ?g]") -- "<b>Parent randy grandpa.</b> <br> ?g = grandpa" --> s(Succeed)
    style s fill:green

The left branch fails because we haven't told the system anything about Sharon's parents.¹  But the right branch succeeds, because the database includes a parent for randy.

Notes