The preregistration of this project can be viewed [here.][1] When people are engaged in a demanding task, unrelated, unexpected objects and events often elude detection, no matter how surprising or odd. Previous work (e.g. Most et al., 2001; Most et al., 2005; Goldstein & Beck, 2016) has shown that whether or not we notice something unexpected has a great deal to do with the objects to which we are attending, the objects we are ignoring, and how simiar or different the unexpected object is to each. Previous inattentional blindness paradigms have traditionaly used objects differentiated by a single feature or category. For example, white and black objects are often used, which are both defined by a single feature (attention sets can beformed to select for white and ignore black, for instance) and a single relationship (lighter and darker). It is thus unknown what the capacity of these sets is and with how much specificity representations of set members can be maintained. In the present study, we define the two sets of objects using multiple features. We use a group of white objects and a group of objects of multiple colors (black, red, purple, and yellow). Thus in order to form a set, either a category boundary must be established (white and nonwhite), or multiple features must be maintained simultaneously (e.g. ignore red things, ignore black things, ignore purple things, and ignore yellow things). These two possibilities yield different predictions for unexpected objects which match the category--that is, are nonwhite--but are a new color that is not an exact match to any instance of that category. If the attention sets are formed along categorical lines, then there should be no difference between a new color and an exact match to an already-present color since both fall into the "nonwhite" category. However, if each member of the set is maintained individually, there should be a difference between an already-present color and a new color. Thus a simple difference score--noticing rates for a nonwhite but uniquely colored objects when attending to white shapes versus nonwhite shapes--will provide evidence for one hypothesis over the other. A difference of 0 is evidence for the set of individuals hypothesis (since in both instances the unique color will be different from what is ignored, and thus noticed at high rates). A negative difference is evidence for the categorical set hypothesis, since the unique color should be suppressed along with other nonwhite objects in the attend white condition (and thus noticed infrequently), and should be noticed at higher rates when attending nonwhite objects. [1]: https://osf.io/8nckf/?view_only=3a0c7972c48744b19e11902e32a73764