Should I avoid instanceof when working with a tree of objects?

Question

I have a class hierarchy representing different language constructs:

Expression <- NumericLiteral
              UnaryExpression
              BinaryExpression
              IndexingExpression
              IteratedExpression
                   ...

The objects of these classes form complex tree hierarchies on which I have to perform various structural checks, e.g. if the node is an IteratedExpression then its first child should be an IndexingExpression. If the check only involves one level I can use the Visitor pattern, but in more complex cases as in the example below I am using instanceof.

void visit(IteratedExpression node) {
    if (!(node.getChild(0) instanceof IndexingExpression)) {
        // report error
    }
}

Is it a proper use of instanceof or I have a flaw in my design? What are the alternatives?

Since there were some alternatives suggested I would like to emphasize the first part of the question:

Is it a proper use of instanceof or I have a flaw in my design?

Answer 1

Like this:

class Visitor {
  boolean indexingExpected;
  void startIteratedExpression() {
    indexingExpected = true;
  }
  void doIndexing() {
    indexingExpected = false;
  }
  void someOtherVisit() {
    if (indexingExpected) {
      throw IllegalStateException();
    }
  }
}
clas IteratedExpression {
  private List<Expression> children;
  public void visit(Visitor visitor) {
    visitor.startIteratedExpression();
    for(Expression child : childrenExpression) {
      child.visit(visitor);
    }
    visitor.endIteratedExpression();
  }
}
class IndexingExpression extends Expression {
  public void visit(Visitor visit) {
    visitor.doIndexing();
  }
}

If you want use Visitor it doesn't matter how many levels you have in your tree.

Answer 2

add an abstract method to Expression and implement it in its children.
so each class will have its own checks.