Partial learning may simply involve attributes (descriptions) of the material being learned all-or-none. An experiment, therefore, biased 41 college student Ss in favor of encoding visual displays in terms of 3 experimentally defined attributes. After Ss reconstructed a group of such visual figures out of a fixed set of component attributes, they were told which attributes had been incorrectly assigned and were given a 2nd guess. Where Ss had indeed encoded the figures in terms of the 3 attributes, 2nd guesses were random. When they did not, 2nd guesses contained information (p = .01). The results are interpreted to mean that attributes are either learned all-or-none, or cross a threshold in an all-or-none fashion. (PsycINFO Database Record (c) 2006 APA, all rights reserved)