Academic journal article Memory & Cognition

Learning Time-Varying Categories

Academic journal article Memory & Cognition

Learning Time-Varying Categories

Article excerpt

Published online: 20 April 2013

© Psychonomic Society, Inc. 2013

Abstract Many kinds of objects and events in our world have a strongly time-dependent quality. However, most theories about concepts and categories either are insensitive to variation over time or treat it as a nuisance factor that produces irrational order effects during learning. In this article, we present two category learning experiments in which we explored peoples' ability to learn categories whose structure is strongly time-dependent. We suggest that order effects in categorization may in part reflect a sensitivity to changing environments, and that understanding dynamically changing concepts is an important part of developing a full account of human categorization.

Keywords Categorization · Memory

(ProQuest: ... denotes formulae omitted.)

At no two moments in time are we presented with the same world: Objects move, plants and animals are born and die, friends come and go, the sun rises and sets, and so on. More abstractly, while some of the rales that describe our worldlike physical laws-are invariant over the course of our everyday experience, others-like legal rules-are not. Given some appropriate time scale, certain characteristics of an entity or class of entities can change; moreover, they may tend to change in systematic ways. For instance, the features that describe phones have changed considerably over recent decades: Not only do modern phones perform many new functions, they are also physically smaller, sleeker, and smoother. Not surprisingly, people's expectations about category members change to suit the environment as it stands: If asked to describe a phone in 2012, few people would refer to a rotary dial, but in 1970, nearly everyone would.

In one sense, nothing is surprising about this observation. However, the changeable nature of many of the concepts and categories with which humans must interact has not been greatly emphasized in the categorization literature (but see Elliott & Anderson, 1995). In category-learning experiments, it is generally assumed that the underlying category is more or less static, and as such, the order in which one encounters category members should not matter to a rational learner. In statistics, this is referred to as the assumption of exchangeability, and for reasons of simplicity, it is generally the default assumption. Current probabilistic models of categorization make this assumption quite explicitly (e.g., Griffiths, Sanborn, Canini, & Navarro, 2008; Sanborn, Griffiths, & Navarro, 2010), and to the extent that standard exemplar and prototype models of categorization can be viewed as kinds of probabilistic models, they can also be seen to abide by this assumption (Ashby & AlfonsoReese, 1995; Griffiths et al., 2008).

Perhaps because exchangeability is assumed in most real-world data analysis, it is generally taken to be a normative standard. However, human learners are also sensitive to the order in which category members are observed; this sensitivity appears to violate this normative standard. One way to account for order effects in cognitive models is to use learning rules that are sensitive to them. Some such rules can be viewed as modifications to standard probabilistic models. For instance, highlighting effects can be accounted for by assuming that people follow a "locally Bayesian" learning rule (Kraschke, 2006), whereas primacy effects can be captured by using a particle filtering learning rule (Sanborn et al., 2010). Another approach is to adopt connectionist, error-driven learning rules, which implicitly assume that recent items are more salient, and so are able to capture some kinds of recency effects, often better than a simple recency-weighting strategy would (e.g., Nosofsky, Kruschke, & McKinley, 1992; Sakamoto, Jones, & Love, 2008). A third approach is to alter the underlying stimulus representation: For instance, certain recency effects can be captured by the assumption that people track the differences between successive observations (Stewart, Brown, & Chater, 2002). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.