We will exclude participants who used one harmfulness rating at least 12 times. The analysis will be conducted using multilevel modeling with participants and additives as random factors. We are primarily interested in the effect of pronounceability on judgment of harmfulness, but we will also use novelty as a secondary measure.
We are mainly interested in interaction of pronounceability and source of the names (original X newly constructed). A significant interaction may suggest that the original effect is limited only to the original words and does not generalize to newly constructed stimuli.