Bilingual false recollection: An EEG study

Abstract

University of Sussex, UKAccepted 6 July 2006IntroductionFalse recollection in the Deese-Roediger-McDermott (DRM) para-digm describes the phenomenon of remembering critical words, e.g.,SLEEP that are thematically related to a list of study words, e.g., BED,REST, AWAKE. Different theoretical accounts of false recollection gen-erate alternative predictions about the false memory effect. Activationmonitoring theory (AMT) (Roediger & McDermott, 1999) assumes thatfalse recollection of critical words results from failure to monitor thesource of activation, which is the same process that is used to correctlyidentify studied words and thus support veridical recollection. FuzzyTrace Theory (FTT) (Brainerd & Reyna, 2002) assumes that false memo-ries arise because participants automatically infer the theme or meaning ofstudied words but confuse words presented at test with nonpresented itemsthat are representative of that theme and veridical recollection depends onverbatim memory. The difference between these accounts is FTT allowsfor dissociable memory traces and AMT does not.ERP studies investigating false recollection report little difference inindices for true and false recollection (Duzel, Yonelinas, Mangun, Heinze,& Tulving, 1997) although Miller, Baratta, Wynveen, and Rosenfeld(2001) reported a shorter P300 latency for false recollection and Fabiani,Stadler, and Wessels (2000) reported less left lateralisation for criticalwords. Other studies, however, report late negativity in false recollectionthat is localised to the frontal lobes (Curran, Schacter, Johnson, & Spinks,2001; Johnson, Kounios, & Nolde, 1996, 1997). No study has examinedfalse recollection in bilingual speakers using EEG. We know bilingualspeakers are slower to access the semantic representations of words pre-sented in their second language (L2) although this effect diminishes withproficiency in L2 (Kroll & Stewart, 1994). It is also known that Span-ish–English bilingual speakers perform recognition tasks in L2 that relyon verbatim memory as well as and in some cases better than they performthe same tasks in L1 (Durgunoglu & Roediger, 1987). One predictionderived from FTT is that bilingual speakers will show reduced gist basedmemory for words presented in L2. By contrast, according to AMT thereis no reason to assume that EEG signals for false and veridical memorywill differ in bilingual speakers. Behavioural data suggests the false mem-ory effect for Spanish–English bilingual speakers is smaller in L2 than L1and veridical recollection is comparable (Anastasi, Rhodes, Marquez, &Vellino, 2005). We predicted that false recollection and veridical recollec-tion in bilingual speakers would reveal a different pattern of EEG signals.Specifically, we expected different patterns in late negativity signals on crit-ical and studied word trials as found by Curran et al. (2001).Subjects and methodsFifteen native Spanish speakers were tested all undergraduates in Sus-sex studying in English. A within participants design was used with Rec-ognition Type (Target words, Critical Distracters, Unrelated Distracters)as the independent variable. Distracters refer to words not studied but pre-sented at test with critical distracters strong associates of study words andunrelated distracters not associated to study words. In the study phase,stimuli were presented in nine blocks with presentation order counterbal-anced. Participants were instructed to remember words in the study lists.Each condition had 36 word lists presented in one of three orders withall participants studying all word lists. The presentation rate was one wordevery 2 s followed by a fixation cross for 300 ms. Participants wereprompted to blink 1 s after each trial. After study, instructions were dis-played for the distracter task. This involved counting backwards in twosaloud from 200 for 1 min. Instructions were then displayed for the recog-nition test. Thirty-nine words were presented for 2 s followed by a fixationcross for 300 ms and a question mark. Participants pressed yes to wordsseen at study and no to words that had not been on the study list. EEGwas recorded from 28 sites using electrodes mounted in a elastic Quikcap(NeuroScan) continuously sampled at AD rate 1000 Hz/channel with ref-erences on left and right mastoids. The vertical electrooculagram (EOG)was recorded with electrodes above and below the left eye. Inter-electrodeimpedance was maintained below 5 kX. EEG and EOG were amplifiedusing a 1–40 Hz band-pass filter. Trials with EOG artefacts (mean EOGvoltage >±150 lV) were excluded from averaging.Results and conclusionsVoltage change during false recollection (i.e., yes responses to criticaldistracters) and during veridical recollection (i.e., yes responses to studiedwords) was calculated. Results from selected electrodes are summarised inFig. 1. ANOVA on the behavioural data found a significant effect of rec-ognition type F(3,14) = 110.98, p < .01, false recollection of criticaldistracters was less likely than recollection of target words and more likelythan recollection of unrelated distracters (p’s < .01). There was also a cor-relation between proficiency in L2 and false recollection r = .54. Inspec-tion of Fig. 1 shows a divergence in waveform amplitude that is greaterthan 1.5 SD between false and veridical trials in the epoch between 600and 700 ms for Fz, Cz and CPz. Note also that the N400 at Fz, Cz anddoi:10.1016/j.bandl.2006.06.099