Relation-Based Categories are Easier to Learn than Feature-Based Categories

摘要: Relation-Based Categories are Easier to Learn than Feature-Based Marc T. Tomlinson (mtomlinson@love.psy.utexas.edu) Bradley C. Love (Love@love.psy.utexas.edu) University of Texas At Austin 1 Station, A8000 Austin, TX 78712 USA Abstract grasping more complex relational matches (Gentner, 1988; Gentner & Ratterman, 1991; Richland, Morrison, Holyoak, 2006), how far does this shift go? Perhaps when confounding factors equated, adults will readily master relation- based concepts feature-based ones. The current study finds that certain relation-based cate- gories acquired and explores the basis for advantage. All four experiments made use stimuli in which featural information supervenes on same perceptual information. Experiment establishes categories comparable categories. 2 rules out advantage arises from a greater difficulty encoding absolute stimulus compared relative 3 suggests faster flexibility af- forded by aligning relations during comparison process is useful completing task. 4 further tests hypothesis using category structure should not benefit power analogical alignment. In case, feature- at rate. Relational reasoning often viewed as pinnacle hu- man intelligence. Accordingly, one common viewpoint learning defined regularities difficult regulari- ties. This view supported developmental trends learn- ing. Studies comparing ing also find feature advantage, but these studies do ground substrate. offers an appropri- ate between learning. Contrary previous studies, we show can be easier. attributable online comparisons memory representation category. Alternative expla- nations difficulties processing vs. ruled out. Keywords: Analogy; Category Learning; Relations Introduction ability grasp hallmark human intelligence (Thompson Oden, 2000). Evaluative such SAT, GRE, or Raven’s Progressive Matrices stress importance thinking. Most research performance supports re- lational advanced competency. Children learn features earlier those 1978). Experts differ novices organizing knowledge within their domain along lines (Chi, Feltovich, Glaser, 1981). Despite work pro- cessing, novel under conditions absent. It possible numerous bal- anced, could prove easier For example, tasks of- ten demands (Kittur, Hummel, 2004; Waltz, Lau, Grewal, Furthermore, no has used rely percep- tual all cases, carrying different. other do- mains, same/different discrimination (Love, Rouder, Wisniewski, 1999) change detection (Kroger, 2004), concordance been achieved, never study. Although many would agree trajec- tory follows progression appreciating Stimulus Design simple scenes were designed specifically address issue whether it harder relationships objects general objects. To avoid vide informative comparison, both over factors. cor- rectly classify scene its features, participants had observe brightness size each object. correctly it, par- ticipants make judgement about size. Each consisted two circles appearing side-by-side. Across trials, varied (small, medium, large) (light, moderate, dark). These combined give overall relation attributes (which side was bigger brighter) over- (size brightness). medium moderate values always manifested once (See Figure 1). purpose investigate either