This is a fairly frequently asked question. For details, see section 7.4.1 and 7.4.3.1 of the specification.
Briefly: a method with a params array is applicable in either its "normal form" or its "expanded form". That is, you can say
PrintLength(new string[] {"hello"}); // normal form
PrintLength("hello"); // expanded form, translated into normal form by compiler.
When given a call that is applicable in both forms, the compiler always chooses the normal form over the expanded form.
Suppose we chose the expanded form every time both were applicable. Suppose you had
void M(params object[] x) {}
How would you actually pass a null array to this thing if we always chose the expanded form? That would be impossible!
Suppose you said
M(new object[] { "hello" });
and we always chose the expanded form. What would this do? Well, an array of objects is an object, so this would choose the expanded form -- it would make another array, wrap this thing up in the array, and pass that!
The choice of expanded form over normal form leads to crazy results. Always choosing the normal form over the expanded form is the more sensible thing to do.