paying attention to attentionin recognitionmemory...

Paying attention to attention in recognition memory:Insights from models and electrophysiology

Chad Dube§, Lisa Payne§, Robert Sekuler§, & Caren M. Rotello†§The Volen Center for Complex Systems, Brandeis University

†University of Massachusetts Amherst

Reliance on remembered facts or events requires memory for their sources, that is, the con-texts in which those facts or events were embedded. Understanding of source retrieval hasbeen stymied by the fact that uncontrolled fluctuations of attention during encoding cancloud important distinctions between competing theoretical accounts. To clarify the issue,we combined electrophysiology (high-density EEG recordings) with computational mod-eling of behavioral results. We manipulated subjects’ attention to an auditory attribute,whether the source of individual study words was a male or female speaker. Posterior alphaband (8-14 Hz) power in subjects’ EEG increased after a cue to ignore the gender of the per-son who was about to speak. With control of subjects’ attention validated by the pattern ofEEG signals, computational modeling showed unequivocally that memory for source (maleor female speaker) reflects a continuous, signal detection process rather than a thresholdrecollection process.

Introduction

Imagine the following scenario. You recently learnedabout a genetically-engineered tomato-tobacco hybridplant, ‘Tomacco.’ Unfortunately, you cannot rememberwhether you learned about Tomacco from The New YorkTimes or from an episode of ‘The Simpsons’. This failureto recall the source of your memory leaves you unsure ofTomacco’s veracity. Clearly, the ability to retrieve sourceinformation is very important. For example, valid infor-mation about the source of a memory can keep us fromcommitting serious faux pas (e.g., relaying the develop-ment of Tomacco as fact), and can also aid recognition ofevents and objects that are encountered in new contexts(Mandler, 1980).

Source memory is commonly assessed using the sourcemonitoring paradigm (Johnson, Hashtroudi, & Lindsay,1993). In this paradigm, subjects study a list of visually-presented words that vary in some contextual detail suchas spatial location or gender of a voice speaking the wordconcurrently. Next, a recognition test is given. Previously-studied and new words are presented sequentially in ran-dom order. For each, subjects make an ‘Old’ or ‘New’ de-cision (item recognition), and follow each ‘Old’ responsewith a decision as to the word’s source (e.g. male or femalevoice; source retrieval). One important aspect of the taskis its combination of traditional recognition memory mea-sures (item recognition) with a cued recall measure (sourceretrieval).

Supported in part by CELEST, an NSF Learning Center (NSFOMA-0835976), NIH T32-NS07292 and AFOSR FA9550-10-1-0420.

Item recognition can be bolstered by recall of source de-tails. This process, termed ‘recollection’ (Parks & Yonelinas,2009), is often assumed to complement a graded feeling ofknowing or ‘familiarity.’ Although most accounts of recog-nition agree that item recognition depends upon a gradedfamiliarity signal, there is ongoing disagreement about thenature of the second, recollection process. The influentialDual-Process Signal Detection model (DPSDT; Yonelinas,1994, 1999) describes recollection as a threshold process:attempts at recollection either do or do not retrieve a de-tail. When recollection succeeds, it is always accurate andalways produces highly-confident ‘Old’ responses. A com-peting model, Unequal-Variance Signal Detection Theory(UVSDT; Mickes, Wais, & Wixted, 2009), assumes all re-sponses reflect a single, continuously-distributed memorystrength variable. That is, UVSDT assumes recollection is acontinuous process.

Importantly, both models focus on operations that oc-cur at retrieval, giving little notice to factors that operateduring encoding. However, recent UVSDT extensions ex-plicitly acknowledge that encoding-related factors are im-portant in the debate (Hautus, Macmillan, & Rotello, 2008).Additionally, assessments of the competing views of recol-lection have often focused on either neuroimaging or cog-nitive modeling (see Malmberg, 2008 for review). Our ap-proach, which combines both methods with an encoding-specific design, allows us to characterize the role of atten-tion in recognition memory and provide a clear assessmentof the nature of recollection.

DPSDT was motivated by a distinctive regularity inratings-based receiver-operating characteristics (ROCs).These functions, which plot the Hit Rate (P("Old"|Old))against the False Alarm Rate (P("Old"|New)) at varying lev-els of response confidence for fixed accuracy, are typically

2 CHAD DUBE§, LISA PAYNE§, ROBERT SEKULER§, & CAREN M. ROTELLO†

Aggregated Memory Strength (Recollection + Familiarity)

New Items Old Items

“Old” “New”

Figure 1. The unequal-variance signal detection model of recog-nition memory (Mickes et al., 2009).

asymmetric for item recognition. Specifically, accuracy atthe leftmost (‘Sure Old’) point on the ROC is higher thansimpler models would predict. DPSDT attributes this re-sult to the selective role played by threshold recollectionin generating accurate, highly-confident ‘Old’ responses(Yonelinas, 1994).

An alternative explanation is that recent exposure in-creases memory strength for old items, and as this increasevaries in magnitude across items, the net result is an in-crease in the variance of old item strength (Starns, Rotello,& Ratcliff, 2012). This explanation forms a key assump-tion of UVSDT, illustrated in Figure 1. According to UVSDT,subjects respond ‘Old’ whenever a test item’s strengthexceeds a response criterion, and respond ‘New’ other-wise. Memory strength is assumed to be continuously-distributed, with higher average strength assigned to olditems. By varying the response criterion’s location (or,analogously, by assuming several criteria are held at once)and plotting the response proportions predicted at each, acurved, asymmetric ROC is generated.

A key point about UVSDT is its conception of recol-lection and familiarity as aggregate strength. That is, re-gardless of the number and nature of processes involvedin making recognition decisions, they are assumed to beadequately described by a single, continuously-distributedmemory strength variable. In other words, it assumes thatrecollection is a continuous process (Mickes et al., 2009;Rotello, Macmillan, & Reeder, 2004).

To evaluate the competing views of recollection, one canassess ROC data for tasks that vary in the degree to which

recollection is required. For source recall, one can plotP("Male"|Male) against P("Male"|Female) at different lev-els of source confidence, for items correctly identified asold. According to DPSDT, the fact that source responsesare collected via cued recall (i.e. only recollection canbe used) means the resulting source ROCs should be lin-ear, as predicted if threshold recollection were operatingalone. Yonelinas (1999) reported the first source ROCs,and concluded the functions were indeed linear, consis-tent with DPSDT. However, more recent work suggests asimpler explanation for source ROC linearity. Specifically,fluctuations in attention to the source dimension duringthe study phase could produce test trials on which littleor no source information is available. The resulting low-accuracy, low-confidence source responses would ‘flatten’the source ROCs (Slotnick & Dodson, 2005). UVSDT ex-tensions incorporating this assumption fit existing sourceROCs well, supporting the attention-fluctuation hypothe-sis (Hautus et al., 2008).

Despite this success, direct empirical investigations ofthe attention-fluctuation hypothesis are lacking: supportfor it has been largely inferred through goodness-of-fit.Fortunately, recent advances in the electrophysiology ofattention suggest novel ways to (i) test the attention-fluctuation hypothesis directly and (ii) track and controlfor fluctuations in analyses of recognition data.

To understand how, we consider electroencephalograph(EEG) oscillations in the alpha (8-14 Hz) frequency band.Recent studies have exploited pre-stimulus alpha-band os-cillations as a marker of selective attention for an upcom-ing stimulus. It is thought that visual attention entailsboth a decrease in pre-stimulus alpha activity over cor-tical regions responsible for actively encoding a stimulus(Thut, Nietzel, Brandt, & Pascual-Leone, 2006), and an in-crease in regions whose function must be reduced to sup-press task-irrelevant processing (Freunberger, Fellinger,Sauseng, Gruber, & Klimesch, 2009; Payne, Guillory, &Sekuler, 2013). These findings have recently been ex-tended into the somatosensory domain, where magne-toencephalographic (MEG) recordings have revealed in-creases (and decreases) in alpha band activity over brainregions linked to ignored (and attended) regions of thebody (Haegens, Luther, & Jensen, 2012; Jones et al., 2010).

Recent work has documented analogous effects in au-dition. For example, Banerjee, Snyder, Molholm, and Foxe(2011) reported a right parieto-occipital electrode topogra-phy for auditory alpha. As they demonstrated, this patterndiffers from those reported in studies of vision, which tendto be most pronounced in centro-occipital electrodes.

Alpha power is an attractive candidate for linking atten-tion at encoding to theories of recognition. However, thegenerality of the auditory effect remains to be established.If Banerjee et al.’s results reflect some general signature ofauditory attention, then a similar effect and topographyshould hold for more complex tasks, like source monitor-ing. Furthermore, such a signature would provide a uniqueopportunity to test the attention-fluctuation hypothesis.Specifically, trials on which high alpha power precedes the

PAYING ATTENTION TO ATTENTION IN RECOGNITION MEMORY:INSIGHTS FROM MODELS AND ELECTROPHYSIOLOGY 3

auditory source stimulus should be associated, during sub-sequent recognition testing, with lower performance andnear-chance source ROCs for those items. Moreover, trialson which pre-stimulus alpha power is low should produceabove-chance source ROCs. If the attention-fluctuationhypothesis is correct, the implications for models of recog-nition are clear: the corresponding source ROCs shouldbe clearly curvilinear, contrary to the results of previousstudies that neither measured nor controlled for atten-tional fluctuations at encoding (e.g., Yonelinas, 1999). Fi-nally, changes in item ROCs with the addition of recollec-tion have been central in motivating DPSDT. Thus it is im-portant to know how the addition of attended source in-formation actually impacts the corresponding item ROCs(Wixted, 2007).

Methods

Subjects.

Eleven subjects participated, eight of whom were fe-male. Subjects’ mean age was 21 years (SD = 1.95). All hadnormal or corrected-to-normal vision as measured withSnellen targets, reported no hearing deficits, and deniedpsychological or neurological disorders. Subjects werenaive to the purpose of the experiment, and were paid forparticipation.

Stimuli.

Singular nouns were presented visually and auditorily.Each had four to eight letters and one to three syllables.Average written frequency was 123 (SD = 117; Kucera &Francis, 1967). Words were displayed in lower case on a21-inch CRT monitor (99.8-Hz refresh rate; screen resolu-tion = 1280£960 pixels). Displays were viewed binocularlyfrom a distance of 57 cm. Audio files were presented at 48db SPL, with average length = 490 ms (SD = 102 ms). Av-erage frequency of the male speaker’s files (computed overfull duration) was 111 Hz (SD = 15.27 Hz); the female’s filesaveraged 235 Hz (SD = 20.33 Hz).

Procedure.

Encoding. Figure 2 illustrates the encoding procedure.Trials were of two types: Attend Voice (AV) and Ignore Voice(IV), with 40 trials of each randomly intermixed. AV trialsbegan with a (300 ms) fixation screen (white font on a blackbackground). A red box followed (500 ms), cuing subjectsto ignore the upcoming stimulus. Boxes were centrally-presented, subtending 0.70± v.a. Next, the trial’s study wordwas presented briefly (100 ms), in either an italic or non-italic font. Pattern masks bookended the stimulus (rows ofXs, 100 ms each). After this, a green box cued subjects toattend the upcoming (male or female) voice, which spokethe study word. A 700 ms blank screen interval onset withthe recording, which offset at a variable period before theblank screen’s offset. A 300 ms fixation screen followed.Subjects were then asked to indicate the voice’s gender viakey-press. RTs > 3000 ms were followed by the words "Too

Slow" (500 ms). Trials ended with the (non-italic) presenta-tion of the study word (3,250 ms). Subjects were instructedto use this time to study the word for an upcoming memorytest.

IV trials resembled AV trials, except that the cues tradedpositions: font presentations were attended and voices ig-nored. Subjects indicated whether the font was italic ornon-italic. As our main goal in this study is to examine theeffects of changes in attention to male and female voices(the source variable explored in key studies of source re-trieval), we fixed the order of the font and voice intervals,relying on font trials to draw attention away from the voiceduring IV trials. Thus, the key interval of each condition forour analyses contains the second attention cue and audi-tory (’source’) stimulus, as emphasized in Figure 2.

Test. All 80 old words, and 40 new words, were randomlypresented.1 Below each was a 6-point confidence ratingscale ranging from 1 ("Sure New") to 6 ("Sure Old"). Sub-jects rated their confidence that each word was old or new.Each old/new scale was followed by a source scale rangingfrom 1 ("Sure Female") to 6 ("Sure Male"). Subjects ratedtheir confidence that the word had originally been spo-ken in a male or female voice. They were instructed thatsource ratings would follow even their new responses, theessential question being "Assuming your ‘New’ decisionwas wrong, what would your best guess be for the voice."This was in keeping with previous experiments that col-lected these responses for theoretical reasons; we do notconsider these responses in our analyses or discuss themfurther.

Analyses

Ratings ROCs were constructed for a) source recall ateach level of attention to source information and b) itemrecognition varying in whether the old items’ source di-mensions were attended at encoding. These functions al-lowed us to (i) assess whether source recollection for at-tended items is best described as a continuous or thresholdprocess, and to (ii) characterize old/new performance inconditions varying in the extent to which recollection con-tributes.

We assessed source ROC linearity by fitting UVSDT toeach function, and comparing its fit to that of a thresh-old model (essentially, DPSDT without the SDT compo-nent) which is consistent with the assumptions of DPSDTas applied to source decisions (Yonelinas, 1999). UVSDTpredicts curvilinear ROCs for above-chance performance,while the threshold model predicts linear source ROCs. Byassessing the relative fit of these two models, conclusionscan be drawn regarding ROC form and the likelihood that

1 One subject served in a version of the experiment that wasidentical to the current protocol, except for the number ofStudy/Old (120) and New words (60). This subject’s data weresimilar to those of other subjects, and excluding the subjectchanged none of our results or conclusions. Thus we report thedata with this subject included.


M or F? dog (study)

+

dog/dog + “dog” (M/F)

It or Not? dog (study)

+

500ms 300ms 500ms 700ms 300ms Untimed 3250ms

dog/dog + “dog” (M/F)

ATTEND voice (Ignore font)

IGNORE voice (Attend font)

300ms

600ms ISI: 300ms 600ms 600ms 500ms 500ms 500ms

Figure 2. Illustration of an encoding trial’s event structure. Each trial began with a fixation cross that oriented the subject to the regionof the computer display within which the trial’s stimuli would be presented. The fixation point was replaced either by a green squareor a red square. The green square cued the subject that the font (italic or non-italic) of the ensuing word should be attended; a redsquare cued the subject to ignore the word. A cue-stimulus interval of 600 ms followed, and then the word was presented. Thereafter,a second cue was presented. This cue was a square that was green if the first cue had been red, or red if the first cue had been green.Following a second 600 ms cue-stimulus interval, an audio clip of the word spoken in either a male or female voice was played. Theauditory presentation was followed by the trial’s probe. Top row: on half the trials, the font display was to be ignored and the auditorypresentation was to be attended. The probe prompted subjects to respond with a keyboard key press whether the voice that spoke theauditory word belonged to a male or to a female. Bottom row: on half of the trials, the font display was to be attended and the auditorypresentation was to be ignored. For these trials, the probe prompted subjects to indicate whether or not the stimulus had been in anitalic font or non-italic font. Every trial ended with a study period during which that trial’s word was visually presented for 3250 ms.

threshold recollection is driving source responses. UVSDTrequires a mean and variance for the ‘male’ source dis-tribution, and 5 ratings criteria. The threshold model re-quires ‘male’ and ‘female’ recollection probabilities, and 5bias parameters. Thus both models use 7 parameters to fitthe 10 independently-varying points of a given attentioncondition. Models were fit using the opti m() procedure inR (R Development Core Team, 2006) to minimize G2

d f =3.

To measure variation in recollection across the itemROCs, we conducted conventional tests of changes in theHit Rate (HR) at a given False Alarm Rate (FAR), as well asfits of DPSDT. For the latter, we first fit a full DPSDT modelto the two item ROCs, allowing PRecol lect i on to vary. Thismodel required 9 parameters: two recollection probabili-ties and two distribution means (for attended and ignoredold items), and one set of five ratings criteria. The fullmodel fit 20 independent datapoints (d f = 11). Its fit wascompared to that of a restricted model where one value ofPRecol lect i on was estimated. The difference in fit, ¢G2, is¬2-distributed with d f equal to the difference in the num-ber of free parameters (i.e. 1).

EEG signals were recorded using a 129-electrode array(Electrical Geodesics Inc.) and high-impedance amplifiers.

All channels were adjusted for scalp impedance < 50k≠.Sensor signals were sampled at 250 Hz with a 0–125 Hzanalogue bandpass filter. Bipolar periocular electrodesrecorded from above and below each eye, and near eachouter canthus.

EEG signals were preprocessed using Matlab’s EEGLABtoolbox (Delorme & Makeig, 2004). Recordings were re-referenced to the grand average. The 0.5-Hz Butterworthhigh-pass and 60-Hz Parks-McClellan notch filters wereapplied. Epochs containing muscle artifacts, blinks, or sac-cades were rejected via independent components analy-sis and visual inspection. Wavelet analysis, cluster permu-tations, and visual display were performed using Matlab’sFieldTrip toolbox (Oostenveld, Fries, Maris, & Schoffelen,2011). Time-frequency representations were computed us-ing Morlet wavelets with width = 4 cycles/wavelet, with 1-70 Hz centers, in 1-Hz steps.

Alpha amplitude was defined by mean oscillatory powerin the band 8–14 Hz. Wavelet alpha power for all electrodeswas calculated for the epoch from cue onset through stim-ulus offset. A cluster-based, non-parametric, randomiza-tion test (Maris & Oostenveld, 2007) compared AV and IV.The cluster-based test statistic was calculated by compar-


0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Ignore Voice, G2df=3 =2.92

P('M'|F)

P('M'|M)

ObservedSDT

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Attend Voice, G2df=3 =1.18

P('M'|F)P('M'|M)

ObservedSDT

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Ignore Voice, G2df=3 =.58

P('M'|F)

P('M'|M)

ObservedThreshold

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Attend Voice, G2df=3 =9.34,p<.05

P('M'|F)

P('M'|M)

ObservedThreshold

P(“Male”|Female)

C: Ignore Voice D: Attend Voice

A: Ignore Voice B: Attend Voice

P(“

Mal

e”|M

ale)

Figure 3. Source recall ROCs varying in whether the source di-mension was attended (green) or ignored (red) at encoding. Fitsof SDT and the linear threshold model show that when source in-formation was attended at encoding (Panels B and D) the subse-quent test data are best described as a continuous SDT process.

ing alpha power for the two conditions at every sensor,leaving the time window open from cue offset to stimu-lus offset. All sensors for which the t-value exceeded the.05 level were clustered via spatial adjacency. The sumof t-values from the cluster with the maximum sum wasthen used as the test statistic, thereby avoiding the prob-lem of multiple comparisons. A reference distribution oftest statistics was generated by randomly permuting thedata across the two conditions 1000 times. A cluster wascharacterized as significant if the proportion of random-ized values larger than the observed test statistic fell belowthe 0.025 level.

Results

Behavioral results

Source results are plotted in Figure 3, conditional onan old/new rating of ‘6’ (‘Sure Old’).2 Performance (in daunits) varies with attention: IV performance (Panels A andC) falls below AV performance (B and D), t (10) = 3.76, p <.01. IV performance is not above chance (i.e. da = 0;t (10) = .27, p = .79). Unsurprisingly, both models fit theIV ROC well: G2

d f =3 = 2.92 and .58, for UVSDT and thethreshold model, respectively. As performance approachesthe minor diagonal, ROCs necessarily become more linear.The key data for assessing the nature of the source vari-able in our study are in the AV condition. These data, plot-ted in Panels B and D of Figure 3, are clearly curvilinear

0.0 0.2 0.4 0.6 0.8 1.00.0

0.2

0.4

0.6

0.8

1.0 Item Recognition

False Alarm RateH

it R

ate

Attend VoiceIgnore Voice

*

Figure 4. Item ROCs varying in whether the contributing olditems’ source dimension was attended (green) or ignored (red) atencoding. Values on the x-axis are repeated at each level of at-tention. The data show that increasing the contribution of sourceinformation to decisions about old/new status of test items pro-duces a significant difference (p < .05) in Hit Rates solely at theleftmost point of the function. This is consistent with the predic-tions of DPSDT, though the source variable producing the differ-ence was best described as continuously-distributed.

and well-described by UVSDT, G2d f =3 = 1.18. The threshold

model, on the other hand, departs significantly from thedata, G2

d f =3 = 1.18, p < .05. These results are inconsistentwith the idea that these source recall responses are drivento a significant extent by threshold recollection. They are,on the contrary, well-described by a continuous model.

It is important to know how item recognition varies withthe addition of this seemingly continuous source infor-mation. The item ROCs, plotted in Figure 4, tell an in-teresting and informative story. Specifically, when sourcerecall is ‘added’ to item responses (green ROC), accu-racy appears to increase primarily at the leftmost, high-confidence ’Old’ operating point. As the FARs of these twofunctions are necessarily equal, we can assess the increaseby comparing HRs. This test confirmed our initial assess-ment: HR is higher for AV than IV at a rating of 6 (leftmostpoint, corresponding to ‘Sure Old’), t (10) = 2.85, p < .05,but fails to reach significance at any subsequent operat-ing points (e.g. t (10) = 1.42, p = .19, for the immediatelyrightward Attend/Ignore pair). In other words, a seem-ingly continuous source variable has produced an effect in

2 Our results did not differ when collapsing over the old/newconfidence levels.


��

��

��

��

�

�

Figure 5. Topographical display of auditory alpha (Ignore Voice >Attend Voice cluster; p<0.01). Electrodes within the cluster show-ing a significant difference between the two test conditions at the0.05 level are indicated by an asterisk (§).

the item ROCs that is similar to that which originally mo-tived DPSDT’s threshold recollection. At the very least, thisdemonstrates that a continuous variable can produce ef-fects that will be misinterpreted as threshold recollection,so long as that continuous variable tends to produce high-confidence ‘Old’ responses.

To assess this implication directly, we fit DPSDT to theitem ROCs in Figure 4. We started with the 9-parameterfull model that allows PRecol lect i on to vary across condi-tions. This model fit well, G2

d f =11 = 8.05, p = .71. Interest-ingly, the best-fitting values of PRecol lect i on differ by .086:PRecol lect i on = .454 and .368 for AV and IV, respectively.That is, DPSDT indicates a nearly 9 percent increase in rec-ollection probability when source recall is ‘added in’. Weevaluated the increase by comparing this model to a re-stricted model in which one value of PRecol lect i on was es-timated. This test showed ¢G2

d f =1 = 3.65, p = .056, indi-cating the increase in ‘threshold recollection’ with addedsource recall was marginally significant.3

EEG Results

Results of the cluster-based permutation test revealed aright-lateralized cluster of 13 posterior electrodes at whichalpha power for the IV condition was higher than for theAV condition (Figure 5). This effect extended across a 200ms epoch following the offset of the cue, but precedingthe onset of the auditory stimulus. Figure 6 shows thetime-frequency transforms averaged across these 13 elec-trodes. Together, the figures reveal that posterior alpha-band activity following a cue to ignore begins immedi-ately upon cue offset and continues throughout the du-ration of the verbally spoken word. The brief alpha ac-tivity at the presentation of both cues and at the onset of

the to-be-attended stimulus likely reflects the well docu-mented alpha-band phase locking that occurs in responseto a stimulus onset (Freunberger et al., 2009).

Discussion

Our behavioral results show that when attention to thesource dimension is controlled at encoding, the resultingsource ROCs are clearly curvilinear and consistent withthe underlying strength distributions of continuous rec-ollection models (Mickes et al., 2009). Further, the re-sults demonstrate that continuous recollection may pro-duce the same increase in high-confidence HR that orig-inally motivated DPSDT’s threshold recollection. Finally,the results demonstrate that DPSDT will falsely interpreta continuous variable as a threshold variable, so long asits contribution largely results in high-confidence ‘Old’ re-sponses. This suggests that DPSDT’s description of recol-lection is not only unreliable, as others have already sug-gested (Rotello, Macmillan, Reeder, & Wong, 2005), but it isalso not valid.

Our study revealed an increase in right-lateralized pos-terior alpha power following cues to ignore an upcomingvoice stimulus. This topographical difference between IVand AV resembles effects of visual attention (Freunbergeret al., 2009; Payne et al., 2013). However, the focus ofcued auditory attention was more lateral than is foundfor vision. The right, parieto-occipital, attention-mediatedcluster reported here is in agreement with previous find-ings (Banerjee et al., 2011), though our quite similar re-sult came from a considerably more complex task. Withinthe frontoparietal network, implicated in top-down con-trol of spatial attention, activation within subsets of re-gions differs between audition and vision (Krumbholz, No-bis, Weatheritt, & Fink, 2009). Evidence from these stud-ies supports the notion of a modality-sensitive attentionalcontrol region. Recent work with nonhuman primates sug-gests this control region may be subserved by cells in theintraparietal sulcus (Grefkes & Fink, 2005).

To date, increases in posterior alpha power have beendemonstrated for to-be-ignored auditory stimuli includ-ing noise bursts (Banerjee et al., 2011) and speech streams(Kerlin, Shahin, & Miller, 2010). To those we add auditorysource stimuli. Our results, which demonstrate the gener-alizability of the alpha signal as a dynamic marker of audi-tory attention, constitute a first step toward a powerful newanalytic strategy for recognition memory. Using the alphasignal, one can track attentional engagement across con-ditions, subjects, and even individual trials (Macdonald,Mathan, & Yeung, 2011). One can then verify the inter-pretation of such EEG oscillations by applying appropri-ate cognitive models whose fits are conditional on the tri-als of interest. Trials with low attention inferred via this

3 A stricter test, holding constant the estimated parametersfor the full model and estimating only PRecol lect i on anew, pro-duced a significantly worse fit in the restricted model, ¢G2

d f =1 =4.78, p < .05.


��

�

��

� ��

�

��

��

��

��

��

�

��

��

��

��

��

��

��

Figure 6. Grand-averaged, time frequency wavelets averaged across the cluster of 13 posterior electrodes for the Attend Voice (leftpanel) and Ignore Voice (right panel) conditions. Cue onset at Time = 0 and voice onset at Time = 1100 are marked by solid white lines.Dashed lines mark the presentations’ offsets.

hybrid approach can then be safely removed or parame-terized. Alpha power could also be treated as a covariateor used to inform the subject and/or item parameters ofhierarchical models (Pratte, Rouder, & Morey, 2010). Ourwork also suggests that in conjunction with model-basedretrieval indices, alpha power can be used to constrain orvalidate the attention parameters of existing and futuremodels (Hautus et al., 2008).

In sum, we showed that attentional fluctuations affectbehavioral measures of item and source memory, and thatsuch fluctuations can distort memory data and result inmisinterpretations by models, like DPSDT, that fail to ac-count for attentional variation. Our results show that mod-els of complex cognitive processes such as recognition andcued recall should follow the lead of recent SDT extensions(Hautus et al., 2008) by accounting for variations in sub-jects’ attention during encoding. The combination of elec-trophysiology with cognitive modeling constitutes a pow-erful tool for improving our understanding of recognitionand attention’s role therein.


References

Banerjee, S., Snyder, A. C., Molholm, S., & Foxe, J. J. (2011). Oscil-latory alpha-band mechanisms and the deployment of spatialattention to anticipated auditory and visual target locations:Supramodal or sensory-specific control mechanisms? Journalof Neuroscience, 31(27), 9923–9932.

DeCarlo, L. (2002). Signal detection theory with finite mixturedistributions: Theoretical developments with applications torecognition memory. Psychological Review, 109(4), 710–721.

Delorme, A., & Makeig, S. (2004). EEGLAB: An open source tool-box for analysis of single-trial EEG dynamics including inde-pendent component analysis. Journal of Neuroscience Meth-ods, 134(1), 9–21.

Freunberger, R., Fellinger, R., Sauseng, P., Gruber, W., & Klimesch,W. (2009). Dissociation between phase-locked and nonphase-locked alpha oscillations in a working memory task. HumanBrain Mapping, 30(10), 3417-3425.

Grefkes, C., & Fink, G. R. (2005). The functional organization ofthe intraparietal sulcus in humans and monkeys. Journal ofAnatomy, 207(1), 3–17.

Haegens, S., Luther, L., & Jensen, O. (2012). Somatosensory antic-ipatory alpha activity increases to suppress distracting input.Journal of Cognitive Neuroscience, 24(3), 677–685.

Hautus, M., Macmillan, N., & Rotello, C. (2008). Toward a com-plete decision model of item and source recognition. Psycho-nomic Bulletin and Review, 15(5), 889–905.

Huang, J., & Sekuler, R. (2010). Attention protects the fidelity ofvisual memory: Behavioral and electrophysiological evidence.Journal of Neuroscience, 30(40), 13461–13471.

Johnson, M., Hashtroudi, S., & Lindsay, S. (1993). Source moni-toring. Psychological Bulletin, 114(1), 3–28.

Jones, S. R., Kerr, C. E., Wan, Q., Pritchett, D. L., Hamalainen, M., &Moore, C. I. (2010). Cued spatial attention drives functionallyrelevant modulation of the mu rhythm in primary somatosen-sory cortex. Journal of Neuroscience, 30(41), 13760–13765.

Kerlin, J. R., Shahin, A. J., & Miller, L. M. (2010). Attentional gaincontrol of ongoing cortical speech representations in a "cock-tail party". The Journal of Neuroscience, 30(2), 620-628.

Krumbholz, K., Nobis, E. A., Weatheritt, R. J., & Fink, G. R. (2009).Executive control of spatial attention shifts in the auditorycompared to the visual modality. Human Brain Mapping,30(5), 1457–1469.

Kucera, H., & Francis, W. (1967). Computational analysis ofpresent-day American English. Providence, RI: Brown Univer-sity Press.

Macdonald, J., Mathan, S., & Yeung, N. (2011). Trial-by-trial vari-ations in subjective attentional state are reflected in ongoingprestimulus EEG alpha oscillations. Frontiers in Psychology, 2,1–16.

Malmberg, K. (2008). Recognition memory: A review of the crit-ical findings and an integrated theory for relating them. Cog-nitive Psychology, 57(4), 335–384.

Mandler, G. (1980). Recognizing: The judgment of previous oc-curence. Psychological Review, 87(3), 252–271.

Maris, E., & Oostenveld, R. (2007). Nonparametric statistical test-ing of EEG- and MEG-data. Journal of Neuroscience Methods,164(1), 177–190.

Mickes, L., Wais, P., & Wixted, J. (2009). Recollection is a continu-ous process. Psychological Science, 20(4), 509–515.

Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2011).FieldTrip: Open source software for advanced analysis of

MEG, EEG, and invasive electrophysiological data. Compu-tational Intelligence and Neuroscience, 2011, 156869.

Parks, C., & Yonelinas, A. (2009). Evidence for a memory thresh-old in second-choice recognition memory responses. Proceed-ings of the National Academy of Sciences of the United States ofAmerica, 106(28), 11515–11519.

Payne, L., Guillory, S., & Sekuler, R. (2013). Attention-modulatedalpha-band oscillations protect against intrusion of irrelevantinformation. Submitted.

Pratte, M., Rouder, J., & Morey, R. (2010). Separating mnemonicprocess from participant and item effects in the assessment ofroc asymmetries. Journal of Experimental Psychology: Learn-ing, Memory, and Cognition, 36(1), 224–232.

R Development Core Team. (2006). R: A language and environ-ment for statistical computing. Vienna: R Foundation for Sta-tistical Computing. http://www.R-project.org.

Rotello, C. M., Macmillan, N., & Reeder, J. (2004). Sum-differencetheory of remembering and knowing: A two-dimensionalsignal-detection model. Psychological Review, 111(3), 588–616.

Rotello, C. M., Macmillan, N., Reeder, J., & Wong, M. (2005).The remember response: Subject to bias, graded, and not aprocess-pure indicator of recollection. Psychonomic Bulletinand Review, 12(5), 865–873.

Slotnick, S., & Dodson, C. (2005). Support for a continuous(single-process) model of recognition memory and sourcememory. Memory and Cognition, 33(1), 151–170.

Starns, J. J., Rotello, C., & Ratcliff, R. (2012). Mixing strongand weak targets provides no evidence against the unequal-variance explanation of zROC slope: A comment on koen andyonelinas (2010). Journal of Experimental Psychology: Learn-ing, Memory, and Cognition, 64(3), 793–801.

Thut, G., Nietzel, A., Brandt, S. A., & Pascual-Leone, A. (2006).Alpha-band electroencephalographic activity over occipitalcortex indexes visuospatial attention bias and predicts visualtarget detection. Journal of Neuroscience, 26(37), 9494–9502.

Wixted, J. (2007). Dual-process theory and signal-detection the-ory of recognition memory. Psychological Review, 114(1), 152–176.

Wu, C.-T., Weissman, D. H., Roberts, K. C., & Woldorff, M. G.(2007). The neural circuitry underlying the executive controlof auditory spatial attention. Brain Research, 1134(1), 187–198.

Yonelinas, A. (1994). Receiver-operating characteristics in recog-nition memory: Evidence for a dual-process model. Journal ofExperimental Psychology: Learning, Memory, and Cognition,20(6), 1341–1354.

Yonelinas, A. (1999). The contribution of recollection and fa-miliarity to recognition and source-memory judgments: A for-mal dual-process model and an analysis of receiver-operatingcharacteristics. Journal of Experimental Psychology: Learning,Memory, and Cognition, 25(6), 1415–1434.

paying attention to attentionin recognitionmemory...

Documents