Eyewitness memory in late bilinguals: evidence for discursive relativity. *.

Abstract

Loftus and Palmer's (1974) ingenious experiments convincingly demonstrated that linguistic changes in the verbal framing of questions result in changes in eyewitness testimonies. Their findings have inspired a new line of research investigating the relationship between language and eyewitness memory. The present study expands the focus of inquiry to bilingual individuals and examines ways in which cross-linguistic differences--and second language learning in adulthood--may influence the participants' performance on memory tasks involving visual recall. The results demonstrate that in instances where availability--or lack--of certain lexically encoded concepts led to differences in narratives elicited from monolingual speakers of Russian and American English, there were also differences between the two bilingual groups. Russians who learned English as a foreign language patterned with monolingual Russians in their recall, while Russians who learned English as a second language used additional interpretive frames, privacy and personal space, available in English but not in Russian. The discussion examines these results from the perspective of discursive relativity, suggesting directions for further study of the relationship between bilingualism, memory, and cognition.

Key words

eyewitness memory

linguistic relativity

Russian-English bilingualism

1 Introduction

In a series of ground-breaking studies, Loftus and Palmer (1974) showed participants several films portraying car accidents and then elicited recalls through differentially worded questionnaires. They found that the form of a question, in their case a change of a single word, can markedly and consistently affect a witness' answer to that question. In particular, the question "About how fast were the cars going when they smashed into each other?" elicited higher estimates of speed than questions that used the verbs collided, bumped, contacted, or hit. Moreover, during a retest a week later, the subjects previously exposed to the verb smashed, were more likely to respond positively to the question whether they saw any broken glass. In explaining their results, the researchers pointed to the fact that the two verbs, hit and smashed, specify different rates of movement, different type of impact, and different likely consequences of the events to which they are referring. Based on the results of the second experiment, involving retest, they also suggested that the verbal label offered subsequently to the event may cause a shift in the mental representation of that event making it more similar to a representation suggested by the verbal label. Their findings and arguments have signaled the start of a new era in memory research, an era which recognized the interaction between language and eyewitness memory (cf. Loftus, 1979; Thompson, Herrmann, Read, Bruce, Payne, & Toglia, 1998).

Given how many bi- and multilinguals regularly come into contact with the justice system, it is all the more surprising to see that almost all of the studies of eyewitness memory to date have been conducted with monolingual participants. The studies of bilingualism in court commonly focus on court interpreting (cf. Berk-Seligson, 1990) and only one study known to this researcher directly addresses bilinguals' eyewitness memory. In this study, Shaw, Garcia, and Robles (1997) asked English speakers and Spanish-English bilinguals to watch a videotaped simulation of a robbery. Then, the participants read a narrative that contained misleading information about the robbery and answered questions about the videotape. The study employed four language combinations for the narrative presentation/memory test: English-English, Spanish-Spanish, English-Spanish, and Spanish-English. It was found that the effects of misleading postevent information were as robust in the cross-language condition as in the same-language condition.

The present study also aims to address eyewitness memory in bilingual individuals--yet from a different perspective. Instead of investigating the effects of postevent misinformation, as is common in research on eyewitness memory, I will examine how cross-linguistic differences may influence bilinguals' initial recalls. In other words, I will consider whether the language of the recall itself has any influence on the contents of the narratives produced by the participants. I will start out by defining the terms used throughout the paper, outlining its conceptual premises, and discussing some studies which offer evidence of the impact of cross-linguistic differences on eyewitness memory. Subsequently, I will present my own study with Russian-English bilinguals where the language of the recall and the context in which the event took place were subject to manipulation. I will end by examining the implications of the findings and outlining directions for future study of the relationship between bilingualism, eyewitness memory, and cognition.

2 Cross-linguistic differences and eyewitness memory: Theoretical framework

Since the focus of the study is on bilingualism, eyewitness memory, and cognition, I begin by defining the three terms as they will be used in the present paper. The term bilingual, in accordance with the scholarly conventions in the field of bilingualism, will refer to speakers of two or more languages, and thus connote both bi- and multilingual individuals. The term cognition will refer to a variety of phenomena which include but are not limited to perception, attention, categorization, inference, reasoning, and socio-cultural belief systems (Lucy, 1997). The terms visual and eyewitness memory will be used interchangeably to discuss participants' recalls of events they either witnessed personally or observed through visual images (pictures, photographs, films, and videotapes). Hunt and Agnoli (1991, p. 381) point out that memory can be based on two different records: "a direct record of the sensory information at the time that we perceive an event and an indirect, linguistically based record of our description of the event to ourselves." In the present paper, I focus on the latter record. In what follows, I discuss three of its linguistic components that may be affected by cross-linguistic differences: (1) lexico-semantically and structurally encoded concepts, (2) scripts, and (3) interpretive frames.

2.1

Lexico-semantically and structurally encoded concepts

Drawing on my previous proposals (Pavlenko, 1999, 2000, in press), I see concepts as mental representations which affect individuals' immediate perception, attention, and recall, and allow members of specific language and culture groups to conduct identification, comprehension, inferencing, and categorization along similar lines. This view acknowledges that concepts are based on both linguistic and perceptual bases and distinguishes between language-based (or language-related) concepts and concepts not immediately linked to language for which speakers of language X may have a mental representation but no specific linguistic means of encoding. Language-based concepts can, in turn, be subdivided into two categories: lexicalized concepts which refer to lexically encoded items, such as natural objects, artifacts, substances, events, or actions, and grammaticized concepts which refer to notions encoded morphosyntactically, such as number, directionality, tense, or aspect (cf. Slobin, 2001).

Recent investigations with monolingual speakers of a variety of languages suggest that cross-linguistic differences in lexically and structurally encoded concepts may impact memory for and thus recall of spatial arrangements (Levinson, 1997; Pederson, Danziger, Wilkins, Levinson, Kita, & Senft, 1998), number of objects (Lucy, 1992b), rate and quality of motion (Slobin, 2000), and body language and emotions (Pavlenko, 2002a). Pederson (1995) also demonstrated that within one linguistic community, different populations may draw on distinct frames for linguistic spatial reference and, consequently, differ systematically in performance on memory tasks. In turn, Arnold and Mills (2001) showed that deaf and hearing signers were superior to nonsigners in memory for such complex and difficult to describe objects as faces and shoes, yet performed the same as hearing nonsigners in memory for easily verbalizable objects (e.g., an apple or a flower).

Together, these studies show that in some cases speakers of different languages may exhibit systematic differences in recall of the same visual stimuli. Notably, however, with the exception of Arnold and Mills (2001), these studies do not suggest that speakers of different languages differ in their ability to remember particular aspects of reality. Rather, they indicate that the presence of certain lexical or structural categories may make certain aspects of reality more salient and verbalizable, and thus easily available for spontaneous recall. In turn, speakers of languages where these categories are absent or concepts are not encoded, may not necessarily pay spontaneous attention to corresponding aspects of reality.

2.2

Scripts

Loftus and Palmer's (1974) work also points to the fact that most, if not all, concepts are linked to scripts, whereby bumped, hit, and smashed involve different rates of movement, distinct types of impact, and different likely consequences of the collision. In the present paper, I will use the terms scripts and schemas interchangeably to refer to hierarchically structured scenarios involving roles and actions, which in turn can be decomposed into further schemas (Fillmore, 1977; Rumelhart, 1980). The field of cognitive psychology accumulated significant evidence that "schematic material dominates other material in accurate recall, in intruded recall, in recognition confidence, in recall clustering, and in resistance to disconfirmation ... Schemata also facilitate inaccurate recall when the information is schema consistent" (Fiske & Linville, 1980, p. 545). In other words, this means that people are much more likely to perceive and internalize information that is relevant to the scripts they possess, to recall schematically-embedded information more quickly and accurately, and to produce distorted accounts where their schematic representations do not correspond to the events perceived (either leaving out important details, or recalling schema-consistent details falsely) (Di Maggio, 1997). Clearly, schemas are culture- rather than language-specific (see also van Hell's paper in this issue). At the same time, linguists argue that the presence of particular lexicalized or grammaticized categorical distinctions, such as hit versus smash, or fall versus flop versus tumble, forces speakers to learn to differentiate among them (Slobin, 2000).

2.3

Interpretive frames

Yet eyewitness testimony is not simply a more or less faithful recall of particular objects and settings. Both psychologists and legal scholars agree that eyewitness recalls are above all narratives and thus subject to narrative conventions (cf. Gewirtz, 1996; Thompson et al., 1998). Narrative conventions refer here to "conventionalized ways of choosing particular elements of the action and setting experienced or seen for inclusion in verbalization (and indeed in memory), and of organizing those events into narratives" (Tannen, 1980, p. 53). Research on cross-linguistic differences in narrative construction shows that speakers of different languages may exhibit systematic differences in what they see as tellable events and in ways they reconstruct these events in stories (Berman & Slobin, 1994; Chafe, 1980; Holmes, 1997; Liebes & Katz, 1990; McCabe & Bliss, 2003; Mistry, 1993; Sherzer, 1987; Slobin, 2000). For instance, while most European languages favor temporal sequencing, narratives told in the American Indian language Kuna focus much more on …