In mathematical speak, it's not the square of the wavefunction that matters, but the square modulus (or square of the absolute value). This is important because the probability of finding a particle in a volume segment of space has to be real. Therefore the integrand has to be real. Squaring a complex function will lead to a complex function, so you must use the absolute value of the function. Multiplying a complex number by its complex conjugate will always yield a real function.
Maybe these links will help,
http://en.wikipedia.org/wiki/Complex_conjugatehttp://en.wikipedia.org/wiki/Absolute_valueIf you don't care about the rigors of it, just always multiple the wavefunction by its complex conjugate. The complex conjugate of a function is just the function with a -i substituted wherever there is an i in the function. If the wavefunction is real, then the complex conjugate is just the function itself - so you're integrating the function squared. In the normalization condition, the two functions being multipled together in the integrand are the same. But do note that in many applications (such as showing two wavefunctions are orthogonal, or, even more so, finding expectation values) the two functions in the integral are not the same, and in these occasions it's important to be sure you're using the complex conjugate of the correct function. Convention is that the first function is always the complex conjugate. That is, we take the integral of ψ
1*[A]ψ
2, where [A] is some operation.