Associative Illusions of Memory

False Memory Research in DRM and Related Tasks

David Gallo

• Publisher: Psychology Press
• Year: 2013
• Topic: Psychology
• eBook ISBN:
  9781134606832
• Subtopic: Cognitive Psychology & Cognition
• Index: Psychology

The last decade has seen a flurry of experimental research into the neurocognitive underpinnings of illusory memories. Using simple materials and tests (e.g., recalling words or pictures), methods such as the famed Deese-Roediger-McDermott (DRM) task have attracted considerable attention. These tasks elicit false memories of nonstudied events that are vivid, long lasting, and difficult to consciously avoid. Additional research shows that these memory illusions are fundamentally related to more complex memory distortions. As a result, this rapidly expanding literature has generated a great deal of excitement - and even some controversy - in contemporary psychology. Associative Illusions of Memory provides an ambitious overview of this research area. Starting with the historical roots and major theoretical trends, this book exhaustively reviews the most recent studies by cognitive psychologists, neuropsychologists, and cognitive neuroscientists. The strengths and limits of various experimental techniques are outlined, and the large body of existing data is meaningfully distilled into a few core theoretical concepts. This book highlights the malleability of memory, as well as the strategies and situations that can help us avoid false memories. Throughout the review, it is argued that these basic memory illusions contribute to a deeper understanding of how human memory works.

PART

I

BACKGROUND

CHAPTER 1

Associations and Errors through History

From Aristotle’s early musings on the mind to computational models of neural networks, the concept of association—or the hypothetical link between mental representations—has reigned supreme in understandings of memory. Of particular relevance to this essay is the role of associations in episodic memory, or the ability to consciously recollect the events or episodes of one’s past (Tulving, 1972, 1983). This form of memory evolved to provide us with a link between present thoughts or environmental cues and relevant information from our past, and this linking power often is attributed to mental associations. The classic view of association, as advocated centuries ago by philosophers such as Thomas Hobbes, John Locke, and James Mill, was based on the intuition that one idea or memory tends to meaningfully lead to another during the course of thought. A common assumption was that the mind forms associations between contiguous events (those close in time or space) and between similar events (those with similar meanings or sensory attributes). At a later point in time, thinking of one event could activate the other via the associative link that was stored in memory.

This philosophical doctrine has always influenced psychology. William James’ (1890) theory of memory, as outlined in his definitive Principles of Psychology, echoed that of the philosophers before him. For James, memory was based on the associative structure of traces in the brain:

The machinery of recall is thus the same as the machinery of association, and the machinery of association, as we know, is nothing but the elementary law of habit in the nerve-centres.… Retention of an experience is, in short, but another name for the possibility of thinking it again, or the tendency to think it again, with its past surroundings. Whatever accidental cue may turn this tendency into actuality, the permanent ground of the tendency itself lies in the organized neural paths by which the cue calls up the experience on the proper occasion, together with its past associates, the sense that the self was there, the belief that it really happened, etc., etc.… These habit-worn paths of association are a clear rendering of what authors mean by ‘predispositions,’ ‘vestiges,’ ‘traces,’ etc., left in the brain by past experience. Most writers leave the nature of these vestiges vague; few think of explicitly assimilating them to channels of association, (pp. 654–655)

Few modern scientists would deny these general assertions, although they could add that cellular long-term potentiation might forge the associations between “nerve-centres,” that specialized brain structures such as the hippocampus seem to associate (or bind together) “experiences” or “past surroundings” that are represented in distributed cortical areas, or that certain regions of prefrontal cortex are involved in making sense of retrieved information, as well as “the sense that the self was there” and “the belief that it really happened.”

In addition to specific episodic memories, associations are relevant to memories for more general knowledge about the world. Each of us carries a large corpus of knowledge in our heads, including specific concepts, abstract ideas, language, and relationships between “things” in the universe. Such general knowledge, or semantic memory, is continually accessed as we try to make sense of the events that we experience, and as we encode episodic memories about these events (see Tulving, 1972). According to many theories, this information is organized and accessed through associative connections. The result can be considered an elaborate associative web of general knowledge about the world and about our particular experiences within it.

The brain relies on associations to accomplish great feats of memory, but like all biological systems the brain is not perfect. Even the healthy brain can make computational errors, leading to a variety of perceptual and cognitive illusions (for an overview see Pohl, 2004). In the same way, associations sometimes can lead our memories astray. Most everyone has had the experience of remembering events (or features of events) that did not occur in a context, but that were in some way associated to events that did occur. Students of memory have described such associative illusions of memory for quite some time, although intensive research into their origins is a relatively recent development.

This chapter provides a brief historical background to the research reviewed in the remaining chapters. Broad historical reviews of associationism in memory already exist (e.g., Anderson & Bower, 1973), as do historical reviews of research on memory illusions and distortions (e.g., Roediger, 1996; Schacter, 1995). Of particular relevance here is the intersection of these two areas: How have different thinkers conceptualized memory errors and associations over time? As reviewed by Roediger, McDermott, and Robinson (1998), associative theories of memory have a long history, but only recently have theorists focused on how associations can cause episodic false memories. Nevertheless, historical trends have greatly impacted current research and modern ideas. References to associative memory errors are nearly as old as the field of psychology itself (Kirkpatrick, 1894), and as experimental methods and theories have changed over time so too have the types of memory errors that were studied as well as the role (and definition) of associations in the resulting theories.

EARLY ASSOCIATIONISM

In his historical review of memory research, Burnham (1889) devoted an entire section to what he called “paramnesia,” referring to “pseudo-reminispanences,” “illusions,” or “hallucinations” of memory. Burnham traced references to illusory memories as far back as Saint Augustine in medieval times, but mostly focused on observations made by influential 19th century thinkers such as James Sully, John Hughlings Jackson, Theodule- Armand Ribot, and Emil Kraepelin. Of particular relevance here was Kraepelin’s idea of “associirende Erinnerungsfälspanhungen”—or associa-tively based memory errors—in which “a present perception calls up by association pseudo-reminispanences of something analogous or related in the past” (Burnham, 1889, p. 433). An example might be walking through a park with a friend and mistakenly remembering that one had done so before, when in fact they had been with a different friend. These and other ideas adeptly presaged modern research on topics such as imagination-based errors, deja vu, source confusions, unconscious attributions, and vague feelings of familiarity. Of course, there were few (if any) experimental studies of these ideas, which were mostly based on introspection, anecdotes, or informal observations of memory abnormalities following brain damage. As Burnham admitted at the outset of his treatise, his examples represented “a class of phenomena long known, yet but little investigated” p. 431).

In fact, at the time of Burnham’s (1889) review, the experimental study of human memory was only a few years in the making. The earliest research on memory, at least by modern standards, often is attributed to Ebbinghaus (1885), who studied the serial learning of lists of nonsense syllables (e.g., DOQ, ZEH, XOT, and so on). Ebbinghaus was primarily interested in the formation and retention of associations, and so he measured how repeated practice (across testing sessions) facilitated the learning of the order of the items in each list. Although his methods had a strong influence on subsequent research, they left little room for the study of pre-existing associations or knowledge on false memories. Indeed, the whole point of using nonsense syllables was to avoid the potential influence of pre-existing associations and meaningfulness on memory performance. (Although, as demonstrated by Glaze (1928) and Hull (1933), even nonwords can be meaningfully processed, so Ebbinghaus’ hope of avoiding pre-existing knowledge in memory experiments went unfulfilled.)

The fact that Ebbinghaus did not focus on false memories does not imply that memory errors for such simple stimuli could not occur. In a classic demonstration of errors in immediate serial recall, Conrad (1964) found that subjects who tried to recall a visually presented string of letters (e.g., M-T-S-C-X-V) often would falsely recall letters with similar sounds (M-T-F-C-X-B). No doubt Ebbinghaus encountered analogous errors in his serial recall experiments. Such errors have been explained by acoustic recoding in short-term memory, so that letters with similar sounds would be confused. This phenomenon provides a good example of how preexisting knowledge associated with incoming stimuli (in this case, their phonemic code) can influence the processing of those stimuli. Even with the simplest stimuli, our brain activates associated information from memory and this process can lead to memory errors.

E. A. Kirkpatrick (1894) published one of the first laboratory demonstrations of memory errors based on meaningful or conceptual associations. In his experiments, Kirkpatrick read a few lists of common objects for subjects to recall. Although his primary interest was in other variables (e.g., age and gender differences on recall), an incidental observation made at the end of his report is of central interest here. He noted that “...when such words as ‘spool,’ ‘thimble,’ ‘knife’ were pronounced [to students] many students at once thought of ‘thread,’ ‘needle,’ ‘fork,’ which are so frequently associated with them. The result was that many gave those words [on the recall test] as belonging to the list” (p. 608). Kirkpatrick’s observations nicely presage the associative response later proposed by Underwood (1965), discussed in the next section, and show how studying one word can lead to falsely remembering a nonstudied associate on a recall test. Kirkpatrick did not present any numeric data to support his observations, but we can assume that false recall did not occur very often in his task—at least not in a systematic way. Kirkpatrick used lists of mostly unrelated words, so that associative intrusions ostensibly would have been idiosyncratic to each subject, depending on their individual associations to the study words. In any event, more rigorous research on the sorts of associative memory errors that Kirkpatrick had noticed did not arise until at least half a century later.

Other early experimental investigations of associative processes in memory can be found throughout the famous works of animal learning theorists, such as Ivan Pavlov, Edward Thorndike, and Clark Hull. A review of these research areas is well beyond the present scope, and the types of associations that were studied are very different from those discussed above in the context of recalling letters or words. It is important to realize, though, that the principles developed from animal learning studies motivated the field of behaviorism that dominated experimental psychology, and associative theories of memory, in the United States in the early 1900s. The central tenet of behaviorism (or at least the extreme version initially advocated by John Watson) was that the past influences behavior through an associative link between a stimulus (S) and a learned response (R). Under this doctrine, known as S→R psychology, researchers primarily were interested in specifying the conditions under which various stimuli could lead to a response, and were resistant to proposing theories regarding the mental processes that could potentially mediate this relationship. Correspondingly, the more “mental” quality of associations (e.g., how pre-existing knowledge and meaning influence processing) was mostly left out of the equation.

Like the Ebbinghaus paradigm, this framework left little room for the study of pre-existing associations on false memories, but the importance of meaningful associations could be demonstrated with these methods too. For instance, a well-known phenomenon in the conditioning literature is that of “stimulus-generalization,” in which a behavior that was trained to occur in response to an arbitrary stimulus (e.g., salivating on hearing a bell, because the bell would be accompanied with food) was then found to occur in response to a perceptually similar but novel stimulus (a buzzer). Importantly, this sort of generalization also occurs for semantically similar stimuli, such as related words in human conditioning experiments (e.g., Lacey & Smith, 1954). Even new stimuli can elicit learned responses, as long as those stimuli activate the trained stimulus via a pre-existing association. An analogous episodic memory phenomenon is false recognition of a new event or word due to its association or similarity to an old one (this effect is discussed later in this chapter). These are very different phenomena, of course, but the common idea is that similar events or stimuli can be mentally confused, eliciting a learned response in one case, and false memory in the other.

Both the Ebbinghaus tradition and behaviorism had strong influences on early research in human memory (or, more appropriately, human “learning” research). In his seminal book, McGeoch (1942) summarized much of the work in the years after Ebbinghaus (1885), which relied on tasks such as nonsense syllable learning, paired-associate learning, and various perceptual-motor activities (e.g., maze learning). Associations continued to dominate theoretical descriptions of memory, including memory errors. For instance, associatively based interference played a prominent role in early research on the learning of simple skills (e.g., maze completion). Many learning errors were conceptualized as “negative transfer,” or the inhibition of new learning by strong associations to an older, irrelevant skill (e.g., the solution to a different maze). The idea of similarity again played an important role, although not necessarily semantic similarity, as the degree of transfer was found to be a function of the perceptual/motor similarity between the two tasks.

The study of “anticipatory errors” is also relevant. Such errors were commonly observed on rote-learning tasks in which the subject would have to master a sequence through repeated learning attempts (e.g., repeated study and serial recall of the same sequence of numbers). Anticipatory errors reflected the fact that subjects sometimes recalled an item earlier in the sequence than it actually had occurred, ostensibly because of an erroneous association formed at an earlier stage in learning. At the time, these types of errors were not considered episodic false memories, but instead were seen as specific instances of the more general principles of learning that equally could apply to motor skills. Errors that mor...

Table of contents

• Cover Page
• Half Title page
• Series Page
• Title Page
• Copyright Page
• Contents
• Preface
• Part I Background
• Part II Basic Theories And Data
• Part III Applications And Data
• References
• Author index
• Subject Index