What is a memory schema? A historical perspective on current neuroscience literature

Highlights

• A historical review reveals schemas are defined by four necessary features.

• They are adaptable, non-specific associative networks, based on multiple events.

• Schemas differ from narratives, categories, gists, and statistical regularities.

• Diverse interpretations of “schemas” in neuroscience studies are compared.

• vmPFC monitors and coordinates neocortical schema activation.

Abstract

The term “schema” has been used to describe vastly different knowledge structures within the memory neuroscience literature. Ambiguous terminology hinders cross-study comparisons and confounds interpretation of the suggested role of the ventromedial prefrontal cortex (vmPFC) in schema functions. Based on an extensive review of the psychological literature, we propose a framework for distinguishing memory schemas from other knowledge structures. The framework includes a definition of schema as possessing four necessary and sufficient features, and four additional features schemas are sensitive to, which are not required but do play a frequent and central role in schema functions. Necessary schema features are (1) an associative network structure, (2) basis on multiple episodes, (3) lack of unit detail, and (4) adaptability. Features schemas are sensitive to are (5) chronological relationships, (6) hierarchical organization, (7) cross-connectivity, and (8) embedded response options. Additionally, we suggest that vmPFC activity observed in studies of schemas corresponds with participants' coordination of existing schemas with ongoing task demands.

Introduction

It has long been observed that prior knowledge, and schema representations in particular, influence memory formation and retrieval (Anderson, 1984, Bartlett, 1932, Carmichael et al., 1932, Craik and Lockhart, 1972, Posner and Keele, 1968). Cognitive neuroscientists have investigated the influences of semantics and knowledge congruency on memory for almost two decades (e.g. Demb et al., 1995, Kapur et al., 1994, Wagner et al., 1998). However, only recently has there been investigation into the neural mechanisms of more complex knowledge structures such as schemas. Specifically, the ventromedial prefrontal cortex (vmPFC), and its interaction with the hippocampus and posterior neocortex, have been implicated in the facilitation of encoding of new information by prior schemas (Tse et al., 2011, Van Kesteren et al., 2010a, Van Kesteren et al., 2010b, Wang et al., 2012).

This renewed interest in the psychological term “schema” has been characterized by heterogeneous usage of the term by cognitive neuroscientists. As applied in the field, schema appears to encompass an assortment of cognitive structures that facilitate encoding, which greatly vary in complexity, organization, and mechanisms. These include the first half of a coherent unique story (Van Kesteren et al., 2010a), multisensory representation of objects (Van Kesteren et al., 2010b), associative rules (Kumaran, Summerfield, Hassabis, & Maguire, 2009), implicit probabilistic statistical regularities (Durrant, Taylor, Cairney, & Lewis, 2011), and in the rat literature: odour-location associative maps (Tse et al., 2007, Tse et al., 2011, Wang et al., 2012). This diversity follows a historical precedent of applying the term loosely to manifold cognitive structures exerting influence over memory encoding and retrieval. Even Bartlett (1932) expressed significant reservations about the use of the term, proclaiming it poorly captured key characteristics of memory schemas as he viewed it, most notably their dynamic nature.

Investigation of the mechanisms by which prior knowledge influences memory formation is hindered by equating diverse cognitive constructs all under the same term. Specifically, the lack of clear definitions obscures the identification of the critical features that make schemas conducive to memory encoding and retrieval. This review will focus on these critical structural features of schemas and will not focus on the function of schemas, which has been addressed by others and does not appear to be subject to dispute (Anderson and Pichert, 1978, Anderson, 1984, Arkes and Freedman, 1984, Bartlett, 1932, Bransford and Johnson, 1972, Carmichael et al., 1932, Cooper et al., 1995, Head and Holmes, 1911, Kumaran et al., 2009, Piaget, 1926, Preston and Eichenbaum, 2013, Rumelhart, 1980, Shea et al., 2008, Tse et al., 2007, Van Kesteren et al., 2013).

As a first step in an attempt to resolve the ambiguity of “schema” and facilitate cross-study communication about prior knowledge influence on memory encoding and retrieval, we review the psychological and educational literature on schemas. Based on this review we propose that schemas possess four necessary features: (1) an associative network structure, (2) basis on multiple episodes, (3) lack of unit detail, and (4) adaptability; and that they also have sensitivity to four additional features: (5) chronological relationships, (6) hierarchical organization, (7) cross-connectivity, and (8) embedded response options. We then map several of the different definitions of schema in recent neuroscience literature onto these features to illustrate their similarities and differences. This comparison, in conjunction with a discussion comparing task demands presumed to rely on schemas, illuminates how studies based on such disparate interpretations of schemas converge in finding an implication of the vmPFC in schema functions.