during a selective attention task, some words are more likely than others to intrude into the unattended channel. according to treisman's attenuation model (with its filter and dictionary), why does this happen?