I agree with the paper Maria shows: The modulation ("roughness" in the case of screams) matters a lot more than the spectrum, or frequency.
Animal communication sounds, which co-evolved with hearing, use pseudo-periodic excitation to be more detectable than random other natural sounds. Since these sounds are also rich in overtones, they tend to use pitches that are low compared to the best frequencies of hearing, and have energy spread across the whole range of frequencies that we hear.