Recollection and familiarity in hippocampal amnesia

Recollection and Familiarity in Hippocampal Amnesia

Patrizia Turriziani,1,2* Laura Serra,2 Lucia Fadda,2,3 Carlo Caltagirone,2,3

and Giovanni Augusto Carlesimo2,3

ABSTRACT: Currently, there is a general agreement that two distinctcognitive operations, recollection and familiarity, contribute to perform-ance on recognition memory tests. However, there is a controversyabout whether recollection and familiarity reflect different memoryprocesses, mediated by distinct neural substrates (dual-process models),or whether they are the expression of memory traces of differentstrength in the context of a unitary declarative memory system (unitary-strength models). Critical in this debate is the status of recognitionmemory in hippocampal amnesia and, in particular, whether the variousstructures in the medial temporal lobe (MTL) contribute differentially tothe recollection and familiarity components of recognition. The presentstudy aimed to explore the relative contribution of recollection and fa-miliarity to recognition of words that had been previously read or thathad been previously generated in a group of severely amnesic patientswith cerebral damage restricted to the hippocampus. A convergent pat-tern of results emerged when we used a subjective-based (remember/know; R/K) and an objective-based (process dissociation procedure;PDP) methods to estimate the contribution of recollection and familiar-ity to recognition performance. In both PDP and R/K procedures,healthy controls disclosed significantly higher recollection estimates forwords that had been anagrammed than for words that had been read.Amnesic patients’ recollection scores were not different for words thathad been generated or that had been read, and the recollection estimatefor words that had been generated was significantly reduced as com-pared to the group of healthy controls. For familiarity, both healthy con-trols and amnesic patients recognized as familiar more words that hadbeen generated than words that had been read, and there was no differ-ence between the two groups. These data support the hypothesis of aspecific role of the hippocampus in recollection processes and suggestthat other components of the MTL (e.g., perirhinal cortex) may be moreinvolved in the process of familiarity. VVC 2008 Wiley-Liss, Inc.

KEY WORDS: amnesia; recognition memory; recollection; familiarity;hippocampus

INTRODUCTION

A convergent pattern of results, obtained across a wide variety of be-havioral studies and measurement methods, supports the idea that twodistinct cognitive operations, recollection and familiarity, contribute torecognition memory. According to Mandler (1980), recollection involvesconsciously remembering the specific experience in which an item was

encountered; conversely, familiarity entails the feelingof a previous encounter with the stimulus without theretrieval of other details (Tulving, 1985; see also Yone-linas, 2002). However, there is a debate about whetherrecollection and familiarity are the expression of dif-ferent memory processes, mediated by distinct neuralcircuits (dual-process models; Aggleton and Brown,1999), or whether they reflect access to memory tracesof different strength in a unitary declarative memorysystem (unitary-strength models; Manns et al., 2003;Wixted and Squire, 2004). Relevant for this debate isthe controversy regarding the role played by the vari-ous structures in the medial temporal lobe (MTL) inrecollection and familiarity (see, for recent reviews,Aggleton et al., 2005; Cipolotti and Bird, 2006; Skin-ner and Fernandes, 2007). Indeed, while those whosupport the dual-process models typically consider thehippocampus critical for recollection and the neocorti-cal areas in the adjacent parahippocampal gyrus (espe-cially, the perirhinal cortex) implicated in familiarityprocesses (Aggleton and Brown, 1999; Brown andAggleton, 2001; Yonelinas et al., 2002; Fortin et al.,2004), those who support the unitary-strength modelsbelieve that there is a lack of specialization in theMTL, with the hippocampus and parahippocampalcortices cooperating with both recollection and famili-arity components of recognition (Manns et al., 2003;Wixted and Squire, 2004).

Most evidence concerning the debate about func-tional specialization of different components of MTLcomes from the investigation of patients with memoryloss associated with selective hippocampal damage(i.e., with the sparing of other MTL structures). How-ever, contradictory findings have emerged from studiesthat directly explored familiarity and recollection com-ponents of recognition memory in these patients.These controversial results may be due to the varietyof modeling approaches adopted to derive quantitativeestimates of the contribution of recollection and fa-miliarity to the overall recognition performance. Onesuch method is the process-dissociation procedure(PDP; Jacoby, 1991), which is based on the premisethat if a subject can recollect a given item then heshould be able to determine when and where it wasinitially studied, whereas familiarity should not sup-port such discrimination. Thus, recollection is consid-ered as the ability to make an accurate relational rec-ognition judgment, and familiarity the conditionalprobability of recognizing that was not recollected.

1Dipartimento di Psicologia, Universita di Palermo, Italy; 2 IRCCS Fon-dazione Santa Lucia, Roma, Italy; 3Clinica Neurologica, Universita diRoma Tor Vergata, Roma, Italy*Correspondence to: Patrizia Turriziani, PhD, Dipartimento di Psicologia,Universita di Palermo, Viale delle Scienze, Ed. 15, 90128 Palermo, Italy.E-mail: [email protected] for publication 6 December 2007DOI 10.1002/hipo.20412Published online 27 February 2008 in Wiley InterScience (www.interscience.wiley.com).

HIPPOCAMPUS 18:469–480 (2008)

VVC 2008 WILEY-LISS, INC.

Using the PDP to analyze cognition performance on a memorytask for both verbal and nonverbal material, Bastin et al.(2004) reported preserved familiarity and impaired recollectionin a patient with selective hippocampal damage because of car-bon monoxide poisoning.

Another modeling method is based on the remember/knowprocedure (R/K; Tulving, 1985) in which subjects are requiredto introspect about the nature of their recognition memoryjudgments and to report whether they recognize items byremembering (i.e., recollection of qualitative information aboutthe study event) or by knowing (i.e., the item is familiar in theabsence of recollection). Then, the probability of a ‘‘remember’’response can be used as an index of recollection, whereas theprobability that an item is familiar is equal to the conditionalprobability that it received a ‘‘know’’ response, given it was notrecollected (Yonelinas and Jacoby, 1995). Mixed results havebeen obtained in studies that used the R/K procedure to inves-tigate the contribution of recollection and familiarity to therecognition performance of amnesic patients, with pathologyrestricted to the hippocampus. Indeed, Manns et al. (2003)reported a comparable decrement of recollection and familiarityin a group of hypoxic patients with neuroradiological evidenceof selective hippocampal atrophy compared to a group ofmatched healthy controls. By contrast, a dissociation betweenintact familiarity but poor recollection has been documented inother studies investigating recognition performance in a groupof patients with mild memory disorder due to hypoxia (Yone-linas et al., 2002) and in individual cases of patients with selec-tive hippocampal damage (Baddeley et al., 2001; Holdstocket al., 2002; Aggleton et al., 2005).

A final method for estimating the relative contribution ofrecollection and familiarity to recognition performance is basedon the receiver operating characteristic (ROC) procedure. Here,patients are required to provide a subjective estimate of confi-dence in their recognition judgments on a continuous scale. Byexamining the effects of varying response criteria in a recogni-tion memory task, an equation is derived that describes howhits and false alarms should be related if performance reflects acombination of recollection and familiarity (Yonelinas et al.,1998, 2002). Inconsistent results have also been obtained instudies that used the ROC procedure to investigate recollectionand familiarity in patients with pure hippocampal amnesia. Infact, while some studies have reported a dissociation betweenimpaired recollection but spared familiarity in the recognitionperformance of these patients (Yonelinas et al., 2002; Aggletonet al., 2005), some others have reported comparable deficits inthe two kinds of processes (Cipolotti et al., 2006; Wais et al.,2006).

The aim of the present study was to explore the relative con-tribution of recollection and familiarity to the recognition per-formance of a group of severely amnesic patients with cerebraldamage restricted to the hippocampus. To compare the resultsderiving from an objective- and a subjective-based measurementmethod, our patients were administered both an R/K and aPDP tasks. Indeed, the main source of discrepant findings inprevious neuropsychological studies may be the variety of mea-

surement methods adopted. In particular, the use of experimen-tal paradigms based on the subjective appreciation of the qual-ity (as in R/K procedure) or of confidence (as in the ROC pro-cedure) in the recognition judgments could be at the origin ofthe variable results in patients, whose ability to introspect theirown memory performances is likely altered. In this regard, itshould be noted that some amnesics have difficulty in subjec-tively appreciating the difference between ‘‘remember’’ and‘‘know’’ (Baddeley et al., 2001; Bastin et al., 2004) associatedwith the problem of maintaining this difference over a test ses-sion. Moreover, following Verfaellie and Treadwell (1993), weexamined recollection and familiarity for words that had beenpreviously read or generated by solving an anagram. In fact,based on a great deal of previous literature on the generationeffect (e.g., Slamecka and Graf, 1978; Verfaellie and Treadwell,1993), we expected that recognition accuracy and, in particular,recollection scores would be higher for real words that hadbeen anagrammed than for words that had been read. However,since previous studies have also documented a reduced genera-tion effect in the memory performance of amnesic patients(Verfaellie and Treadwell, 1993), we expected that impairedrecollection in patients with amnesia would be particularlysevere for words that had been generated.

MATERIALS AND METHODS

Subjects

Four amnesic patients (all men) with damage limited to thehippocampal formation participated in the present study.Strictly behavioral and neuropsychological criteria were adoptedto select patients for this group. In particular, all of thesepatients were referred to our laboratory because of impairmentin recalling day-to-day events, and their neuropsychologicalevaluations showed a severe long-term episodic memory deficitin the context of substantially preserved general cognition. Thefirst patient, TE, developed amnesia in 1996 following carbonmonoxide poisoning. The second, CM, an insulin-dependentdiabetic, became amnesic in 1998 after a period of comabecause of iatrogenically caused hypoglycemia. The last twopatients, AR and CC, became amnesic in 2000 and 2003,respectively, following a drug (heroine) overdose and associatedrespiratory failure. Table 1 summarizes the biographical andclinical characteristics of the amnesic patients as well as per-formance scores on some neuropsychological tests. All patientshad medium- to high-range WAIS verbal IQs, performed inthe range of normal controls on tests of visuoconstructive abil-ities and deductive reasoning and, with the exception of CC,achieved a normal score on a test of executive functioning (themodified version of the Wisconsin Card Sorting test). All ofthem, however, obtained less than normal scores on tests ofverbal and visuospatial episodic memory. The mean age of thepatients in the amnesic group was 49.6 (614.5) yr and averageeducation was 13.8 (61.8) yr.

470 TURRIZIANI ET AL.

Hippocampus

A group of eight healthy volunteers, matched for age and lit-eracy with the patients, served as controls for performance onthe neuropsychological tests assessing the recollection and fa-miliarity components of recognition memory. The mean ageand education of this control group were 53.3 (613.1) yr and14.5 (63.0) yr, respectively. Moreover, to evaluate results ofthe neuroradiological examination, the data of each amnesicpatient were compared with those of a group of five healthyvolunteers (overall 20 subjects) matched for age (48.8 6 9.2)and gender (all males).

Exclusion criteria for subjects in the healthy control groupswere the following: a history of head trauma, anamnestic orcurrent neurological or psychiatric diseases, severe systemic dis-eases, or taking drugs or medications potentially able to inter-fere with cognitive efficiency.

Neuroradiological Investigation

Visual inspection of the four patients’ clinical MR images ofthe brain revealed bilateral atrophy of the hippocampus; nofocal lesions or atrophic changes were evident in the nonhippo-campal structures of the MTL or in the rest of the brain paren-chyma outside the MTL. To obtain a morphovolumetric mea-surement of the hippocampal formation and parahippocampalcortices, we made a manual segmentation of the MTL struc-tures with a 3D data acquisition protocol. Magnetization pre-pared rapid gradient echo (MPRAGE) T1-weighted images(Erlangen, Germany; 1 mm isotropic voxel, TR 11.4 ls, TE4.4 ls, flip angle 5 158) were obtained with a Siemens 1.5 TVision Magnetom MR system. To reduce the interindividualvariability in gross brain size and to circumvent the error-pronecollection of an index of brain size, each of the brain volumesfrom which the volumetric measures had been collected wereregistered into the Talairach proportional stereotaxic spaceusing a nine-parameter registration algorithm (Collins et al.,

1994). Anatomical landmarks used for the manual segmenta-tion of the hippocampi and the perirhinal, entorhinal, and par-ahippocampal cortices in the parahippocampal gyrus were inaccordance with Insausti et al. (1998) and Pruessner et al.(2000, 2002). The ROIs were mapped using the interactiveprogram DISPLAY (McDonald JD, Brain Imaging Center,Montreal Neurological Institute;www.bic.mni.mcgill.ca/soft-ware/Display).

In order to exclude that, as a result of anoxia, the amnesicpatients presented diffuse neocortical injury as well as MTLinvolvement, a whole brain morphovolumetric analysis was per-formed. For this purpose, the three-dimensional T1-MPRAGEimages of patients and controls were processed using SPM 2(Wellcome Department of Cognitive Neurology, London, UK;ttp://www.fil.ion.ucl.ac.uk/spm). According to the optimizedvoxel-based morphometry approach (Good et al., 2001), thenative gray matter map was normalized to customized tem-plates. The optimized transformation parameters were thenapplied to the original T1 volumes in native space. Finally,these normalized images were segmented to produce gray mat-ter, white matter, and cerebrospinal fluid maps in normalizedspace. For each of the amnesic patients and controls, the wholebrain parenchyma volume was calculated as the sum of thegray and white matter volumes. The control group was scannedusing the same protocol on the same scanner as the patients.Moreover, the same individual made the volumetric measuresof both patients and controls. The same group of healthy con-trols was used to compare the volumes of both the MTL struc-tures and the whole brain.

Table 2 reports the volumes of the MTL structures and ofthe whole brain parenchyma in the four amnesic patients. Thehippocampi were consistently and bilaterally atrophic, and thevolume reduction ranged from 1.98 to 4.81 standard deviationsbelow the mean of the age- and gender-matched healthy con-trols. The hippocampal atrophy was particularly severe in TE

TABLE 1.

Performance Scores on Some Psychometric and Neuropsychological Tests of the Amnesic Patients Who Participated in the Present Study

Patients

Age

(yr) Sex

Education

(yr)

Etiology

of cerebral

damage WAIS-VIQa

Raven’s

progressive

matrices

Rey’s

figure

copy

Modified WCST

(number of

categories)b

Delayed

word-list

recallc

Delayed

prose

recalld

Delayed

Rey’s figure

reproductione

TE 50 M 13 Hypoxia 110 28.3 16.6 6 0f 2.7f 2.6f

CM 47 M 13 Hypoglycaemic

coma

128 31.2 35 5 0f 3.1f 3.0f

AR 30 M 17 Hypoxia 115 25.3 27.6 6 0f 2.1f 4.1f

CC 44 M 13 Hypoxia 95 30.5 20.9 2f 0f 0f 0f

Superscript alphabets a–e and g provide, for each test, references to the original work as well as to the paper presenting normative data for the Italian population.aWechsler, 1981; Orsini and Laicardi, 1997.bNelson, 1976; Nocentini et al., 2002.cRey, 1958; Carlesimo et al., 1996.dSpinnler and Tognoni, 1987; Carlesimo et al., 2002.eRey, 1941; Carlesimo et al., 2002.fIndicates a pathological score (below the lower limit of 95% tolerance interval measured in the normal population).gRaven, 1949; Carlesimo et al., 1996.

RECOLLECTION AND FAMILIARITY IN THE HIPPOCAMPUS 471

Hippocampus

and CM, less pronounced in AR, and of intermediate severityin CC. There was no indication of significant volume reduc-tion in any of the parahippocampal cortices in TE, AR, andCC. CM presented volumes in the range of normal controlsfor the perirhinal cortex bilaterally and for the left entorhinaland right parahippocampal cortices. However, the volumes ofhis right entorhinal and left parahippocampal cortices wereabout two standard deviations below the mean of the controlgroup. It should be noted that a similar pattern of atrophicchanges in the parahippocampal gyrus was also obtained whencortical volumes were corrected for the variability of the collat-eral sulcus, according to Pruessner et al. (2002). Finally, therewas no evidence of diffuse neocortical injury. Indeed, the wholebrain volume was substantially normal in CM and CC andonly slightly reduced in TE and AR.

Process Dissociation Procedure

Materials

Stimuli for this experiment consisted of 160 five-letterwords, with a frequency of occurrence in the Italian languageranging from 2 to 331, with a mean of 27.1 per million (Bor-tolini et al., 1971). The 160 words were divided into two 80-word sets that were comparable for frequency of occurrence.The first set was used to create the study and test lists for theinclusion condition, whereas the other set was used to createthe study and test lists for the exclusion condition. The assign-ment of lists to conditions was counterbalanced across subjects.For each of the two sets, 25 words were selected as stimuli tobe presented visually during the study phase. Thirteen of thesewere designed to be read aloud by the subject, and the remain-ing 12 were presented in an anagram format. Anagrams werecreated by underlining the second and fourth letters of a wordand keeping them in their correct positions. Two of theremaining three letters were randomly reversed (e.g., loplo 5pollo). Using these criteria, the target word became the onlypossible solution for each anagram. Next, 30 words wereselected to be presented auditorily during the study phase. Therecognition test contained all 25 visually presented words, all30 auditorily presented words and 25 unstudied words.

Procedure

In order to become familiarized with the anagram task, thepatients and control subjects first solved five practice anagrams.The experiment began when the patients had solved the prac-tice items and fully understood the task. The study phase wasidentical for both inclusion and exclusion conditions. Wordswere presented visually or auditorily in two distinct blocks.The patients were first told that a series of words would be pre-sented on a PC screen for a subsequent memory test and thatsome of the words would be anagrams they should try to solveand some words in normal form they should simply readaloud. If the subjects solved an anagram incorrectly, they wereencouraged to continue trying to find the correct solution for amaximum of 1 min. At this point, if they were unsuccessful,T

ABLE2.

VolumetricData

oftheMTLStructuresandoftheW

hole

Brain

intheAmnesic

Patients

WhoParticipatedin

thePresentStudy

Patients

Hippocampus(m

m3)

Perirhinal

cortex

(mm

3)

Entorhinal

cortex

(mm

3)

Parah

ippocampal

cortex

(mm

3)

Whole

brain

(grayan

dwhitematter)

(liters)

Right

Left

Right

Left

Right

Left

Right

Left

TE

1,92

7(2

3.62)

2,08

1(2

4.81

)1,918(2

0.03)

2,258(1

1.07)

1,91

5(1

0.78)

2,49

9(1

1.79

)2,071(2

0.25)

1,938(2

1.05

)0.991(2

1.43)

CM

2,17

8(2

3.24)

2,30

3(2

4.23

)1,657(2

1.04)

1,957(2

0.05)

1,22

2(2

1.95)

1,75

9(1

0.03

)1,848(2

0.74)

1,533(2

2.41

)1.116(2

0.31)

AR

3,32

4(2

1.98)

3,38

9(2

2.05

)3,223(1

2.00)

2,228(2

0.26)

2,51

9(1

2.24)

2,78

4(1

2.82

)2,394(2

0.05)

2,509(1

0.21

)1.116(2

1.68)

CC

2,79

4(2

2.23)

2,75

5(2

3.15

)2,121(1

0.76)

3,512(1

2.74)

1,92

0(1

0.80)

1,94

5(1

0.47

)2,033(2

0.33)

2,128(2

0.41

)1.120(2

0.28)

The

numberof

standarddeviation

saboveor

belowthemeanof

thehealthycontrols(n

55foreach

oftheam

nesicpatients)isgivenin

parentheses.

472 TURRIZIANI ET AL.

Hippocampus

the examiner provided the solution. During this phase, subjectswere occasionally reminded that the second and fourth lettersof the anagrams did not need to be changed. When the visualpresentation was completed, the auditory presentation beganimmediately. Words were spoken aloud by the examiner, andthe patients were told that they should repeat each word andtry to remember it for a subsequent memory test.

During the recognition phase, the 80 words were presentedon a PC screen. For each word, the patients had to decidewhether it was old or new. However, instructions differed forthe inclusion and exclusion conditions. In the inclusion condi-tion, the patients were instructed to call an item ‘‘old’’ if it hadbeen visually presented or heard earlier and ‘‘new’’ if it had notbeen presented before. In the exclusion condition, patients wereinstructed to call an item ‘‘old’’ only if they had heard it duringthe study phase, and ‘‘new’’ only if they had previously read itor given it as a solution to an anagram, or if it had not beenpresented before.

The inclusion and exclusion conditions of the experimentwere administered to each patient on the same day. Care wastaken to emphasize that the instructions for the recognition testwere different in the two conditions.

Remember–Know Procedure

Materials

Stimuli for this experiment consisted of 80 five-letter words,with a frequency of occurrence in the Italian language rangingfrom 1 to 83 with a mean of 27.3 per million (Bortolini et al.,1971). Twenty-five words were selected for visual presentationduring the study phase. Thirteen of these had to be read aloudby the subject, and the remaining 12 were presented in an ana-gram format. Next, 30 stimuli were selected for auditory pre-sentation during the study phase. Finally, the remaining 25stimuli served as foils for the recognition test. The recognitiontest contained the 25 words presented visually, the 30 wordspresented auditorily during the study phase, and the 25unstudied words.

Procedure

The study phase was identical to that used in the PDP pro-cedure. During the recognition test, the 80 words were pre-sented on a PC screen and, for each one, the patients wereasked to respond ‘‘remember,’’ ‘‘know,’’ or ‘‘new.’’ They weretold to respond remember if they could recollect the specificepisode during which the word had been previously presented.They were told to respond ‘‘know’’ if they were sure the itemhad been presented during the study phase but could not recol-lect the specific episode during which it had been previouslypresented. The subjects were told to respond ‘‘new’’ if theythought the item had not been presented during the studyphase. For several of the test items, participants had to explainwhy they had made a particular response to ensure that theyunderstood the R/K distinction. None of the participantsappeared to have any difficulty understanding the instructions.

RESULTS

Process Dissociation Procedure

Both patients and healthy participants were fast and accurateat solving anagrams during the study phase. The proportions ofcorrectly solved anagrams were 0.98 and 0.97 in the twogroups. The proportions of ‘‘old’’ responses made by the amne-sics and controls in the inclusion and exclusion conditions ofthe test are reported in Table 3. In the inclusion condition, theprobability of calling an item ‘‘old’’ reflects hit rates for studiedwords and false alarm rates for new words. In the exclusioncondition, it reflects hit rates for auditorily studied words andfalse alarm rates for anagrams, read words, and new words. Asfor the auditorily presented words, no difference emergedbetween amnesics and healthy controls either in the inclusion(t 5 1.49; P 5 n.s.) or the exclusion (t 5 1.34; P 5 n.s.) con-ditions. Moreover, both patients and controls made a compara-ble number of hit rates in the inclusion and exclusion condi-tions [t(3) 5 0.10 and t(7) 5 0.71 in the amnesic and controlgroups, respectively; P 5 n.s. in both cases]. As for the visuallypresented words, an analysis of variance (ANOVA) with group(amnesics vs. controls) as between-subject factor and condition(read vs. anagrammed vs. new words) as within-subject factoron the proportion of ‘‘old’’ responses in the inclusion conditionshowed only a significant condition effect [F(2, 20) 5 44.45,P < 0.001] due to the higher rate of ‘‘old’’ responses for wordsthat had been anagrammed (0.71) than for words that hadbeen read [0.54; F(1, 10) 5 9.1, P < 0.01] which, in turn,were classified as old more frequently than unstudied words[0.13; F(1, 10) 5 36.02, P < 0.001]. Neither the group effect[F(1, 10) 5 0.45, P 5 n.s.] nor the group 3 condition inter-action [F(2, 20) 5 0.94, P 5 n.s.] was significant. An analo-gous ANOVA applied to the proportion of ‘‘old’’ responses inthe exclusion condition revealed, instead, a significant condition

TABLE 3.

Average Proportion (and SD) of ‘‘Old’’ Responses to Words That

Had Been Read, to Words That Had Been Produced as a Solution

for an Anagram and to Unstudied Words Made by the Amnesic

and Control Groups in the Inclusion and Exclusion Conditions

of the Process Dissociation Procedure

Type of word Inclusion Exclusion

Amnesics

Auditory 0.70 (0.09) 0.71 (0.21)

Read 0.62 (0.29) 0.35 (0.28)

Anagram 0.71 (0.33) 0.59 (0.28)

New 0.15 (0.06) 0.16 (0.12)

Controls

Auditory 0.59 (0.14) 0.55 (0.18)

Read 0.46 (0.23) 0.16 (0.14)

Anagram 0.72 (0.12) 0.19 (0.14)

New 0.11 (0.10) 0.12 (0.17)

RECOLLECTION AND FAMILIARITY IN THE HIPPOCAMPUS 473

Hippocampus

effect [F(2, 20) 5 21.95, P < 0.001] and group 3 conditioninteraction [F(2, 20) 5 12.45, P < 0.001]. Planned compari-sons revealed that also in this case words that had been ana-grammed were classified as old more frequently than wordsthat had been read [0.38 vs. 26; F(1, 10) 5 25.70, P < 0.001]and these more frequently than unstudied words [0.14; F(1,10) 5 8.84, P < 0.01]. Moreover, amnesic patients mademore frequent incorrect ‘‘old’’ responses than healthy controlson the anagrammed words [F(1, 10) 5 11.27, P < 0.01] butnot on the read [F(1, 10) 5 2.56, P 5 n.s.] and the unstudiedones [F(1, 10) 5 0.13, P 5 n.s.].

We estimated the contribution of familiarity and recollectionto recognition performance of amnesics and controls using thedual-process signal detection model (Jacoby et al., 1993; Yone-linas et al., 1998). In this model, familiarity is assumed toreflect a signal detection process of the type that underliesstandard d 0 tables. Instead, recollection is an independentthreshold process. Thus, recollection is measured as the differ-ence between corrected recognition performance in the inclu-sion and exclusion conditions, each computed by subtractingthe proportion of false-alarm responses from the hit-rateresponses. Instead, false alarms arise because the familiarity ofsome of the new items exceeds the participants’ response crite-rion. In this way, familiarity is measured as the differencebetween the average familiarity of the old and new items (i.e.,d 0, see Yonelinas et al., 1998, for a more detailed description ofthe model). In particular, in the inclusion condition, the sub-jects would correctly respond ‘‘yes’’ to the previously seen words(either read or anagrammed) either because they recollectedthem or because they felt the words were familiar even without

recollecting anything about them (F[1 2 R]). Thus, the proba-bility of responding ‘‘yes’’ in the inclusion condition (Inc) wasgiven by the formula: Inc 5 R 1 F[1 2 R]. In the exclusioncondition, the subjects would incorrectly respond ‘‘yes’’ to apreviously seen word if they found it familiar and failed to rec-ollect the circumstances of its prior presentation. Therefore, theprobability of responding ‘‘yes’’ in the exclusion condition(Exc) corresponded to: Exc 5 F[1 2 R]. In order to obtain anestimate of the recollection process on performance, the proba-bility of responding ‘‘yes’’ in the exclusion condition was sub-tracted from the probability of responding ‘‘yes’’ in the inclusioncondition (R 5 Inc 2 Exc). Finally, an estimate of the effect offamiliarity was obtained by computing: F 5 Exc/[1 2 R].

Figure 1 shows the parameter estimates based on the inclu-sion and exclusion scores using the dual-process signal-detec-tion model. This was performed separately for anagrams andfor words read during the study phase. On the basis of theseestimates, we analyzed separately the contribution of recollec-tion and familiarity to recognition memory. An ANOVA withgroup (amnesics vs. controls) as between-subject factor andcondition (read vs. anagram) as within-subject factor performedon the recollection scores revealed nonsignificant group [F(1,10) 5 1.17, P 5 n.s.] and condition [F(1, 10) 5 0.16, P 5n.s.] main effects but a significant group 3 condition interac-tion [F(1, 10) 5 10.01, P < 0.01]. Planned comparisons madeto qualify this interaction indicated that the amnesics’ meanrecollection score was significantly lower than that of the con-trols only for anagrams [F(1, 10) 5 7.20, P < 0.01], whereasfor read words amnesics’, the mean recollection score did notdiffer from that of controls [F(1, 10) 5 0.15, P 5 n.s.]. Also,

FIGURE 1. Average probability estimates of recollection andfamiliarity for amnesics and controls from the process dissociationprocedure. Bars represent 1 SD from the mean.

FIGURE 2. Average probability estimates of recollection and fa-miliarity for amnesics and controls from the remember–know experi-ment. Bars represent 1 SD from the mean.

474 TURRIZIANI ET AL.

Hippocampus

the controls were more likely to recollect anagrams than readwords [F(1, 10) 5 9.51, P < 0.01], but this was not the casefor the amnesic patients [F(1, 10) 5 2.86, P 5 n.s.]. An analo-gous ANOVA performed on the familiarity scores showed a sig-nificant effect of condition [F(1, 10) 5 16.75, P < 0.001],indicating that anagrams tended to be recognized on the basisof familiarity more frequently than read words. The effects ofgroup [F(1, 13) 5 1.88, P 5 n.s.] and the group 3 conditioninteraction [F(1, 10) 5 0.12, P 5 n.s.] were not significant,indicating that amnesics and controls relied on familiarityequally in making their recognition judgments.

As recollection scores (at least for the anagrammed words)were significantly higher for healthy controls than for theamnesic patients, one wonders whether the comparable famili-arity scores in the two groups are actually a true effect or,instead, the result of a floor effect in the estimates of familiarityfor the healthy controls, because of the lower number of nonre-collected words (i.e., words that could be recognized based onfamiliarity). This is unlikely for two reasons. First, according toYonelinas et al. (1998), we computed the estimates of familiar-ity as the ratio between the probability of incorrectly respond-ing ‘‘old’’ to a previously seen word in the exclusion conditionand the number of nonrecollected words, thus correcting therough familiarity scores for the number of words that couldactually be recognized based on familiarity. Second, there is noevidence in the performance of the healthy group that the scoredistribution for the familiarity estimates was more left-skewedthan that relative to the recollection estimates. In fact, the Kol-mogorov-Smirnov test indicated that in no case did score distri-bution significantly depart from normality (d 0 ranged from0.16 to 0.22 across recollection and familiarity scores; P 5 n.s.in all cases). Moreover, the number of healthy controls whosescores fell more than 1 standard deviation (SD) below thegroup mean was similar for the recollection and familiarity esti-mates (n 5 2, 1, 2, 1 for the recollection and familiarity esti-mates of read and anagrammed words, respectively).

The qualitative performance pattern that emerged from thegroup comparison (i.e., particularly poor recollection of ana-grammed words in the amnesic patients) was also confirmed byanalyzing individual patients’ performances. Table 4 reports the

estimates of recollection and familiarity derived from the PDPfor each amnesic patient. The relatively high variation coeffi-cients in the score distribution of the group of healthy controls(ranging from 0.32 to 0.87 for the recollection and familiarityestimates) prevented us from making a formal statistical analy-sis of individual performances. However, setting a qualitativelimit of average performance at 1 SD below the mean of thehealthy controls, we noted that two out of the four amnesicpatients (TE and CC) obtained below average recollectionscores for the anagrammed words and one patient (AR)obtained a recollection score below the average of normal con-trols for both read and anagrammed words. By contrast, onlyone patient (AR) obtained a familiarity score below the averageof healthy controls for the anagrammed words.

Remember–Know Procedure

Also in this case, during the study phase, patients andhealthy controls made very few errors in solving the anagrams(proportion of correct responses 0.99 and 0.96, respectively).Table 5 presents the proportion of remember and knowresponses for words previously listened to, read, or solved asanagrams and for the unstudied words (false alarms). AnANOVA with group (amnesics vs. controls) as between-subjectfactor and condition (read vs. anagrammed vs. new words) aswithin-subject factor on the proportion of ‘‘remember’’responses showed a significant condition effect [F(2, 20) 511.26, P < 0.001] and a trend toward significance for thegroup 3 condition interaction [F(2, 20) 5 3.20, P 5 0.06].Planned comparisons revealed that, overall, anagrammed wordsreceived more remember responses than read words [F(1, 10)5 3.76, P 5 0.07] and these, in turn, more than unstudiedwords [F(1, 10) 5 10.49, P < 0.01]. This, however, was onlytrue for the healthy controls. Indeed, these patients producedmore ‘‘remember’’ responses on the anagrammed than on theread words [F(1, 10) 5 12.78, P < 0.01] and on the readthan on the unstudied words [F(1, 10) 5 8.21, P < 0.01].

TABLE 4.

Estimates of Recollection and Familiarity for Each Amnesic Patient

Derived From the Performance on the Process Dissociation Procedure

Recollection (probability) Familiarity (d0)

Read Anagrammed Read Anagrammed

TE 0.22 (20.11) 0.14 (21.43) 0.94 (10.44) 2.50 (13.10)

CM 0.78 (12.67) 0.26 (20.89) 1.06 (10.64) 1.20 (20.13)

AR 0.00 (21.22) 0.00 (22.04) 0.53 (20.25) 0.78 (21.18)

CC 0.21 (20.18) 0.15 (21.39) 1.73 (11.77) 2.49 (13.09)

The number of standard deviations above or below the mean of healthy subjectsis given in parentheses.

TABLE 5.

Average Proportion (and SD) of ‘‘Remember’’ and ‘‘Know’’ Responses

to Words That Had Been Read, to Words That Had Been Produced

as a Solution for an Anagram and to Unstudied Words Made by the

Amnesics and Controls in the Remember–Know Procedure

Type of word Remember Know

Amnesics

Auditory 0.53 (0.23) 0.15 (0.08)

Read 0.29 (60.25) 0.15 (60.11)

Anagram 0.27 (60.17) 0.27 (60.12)

New 0.12 (60.14) 0.11 (60.09)

Controls

Auditory 0.46 (0.22) 0.14 (0.06)

Read 0.24 (60.23) 0.21 (60.06)

Anagram 0.48 (60.27) 0.17 (60.15)

New 0.06 (0.08) 0.09 (60.10)

RECOLLECTION AND FAMILIARITY IN THE HIPPOCAMPUS 475

Hippocampus

Instead, in the group of amnesic patients, only a trend towardsignificance was observed in the comparison of read andunstudied words [F(1, 10) 5 3.76, P 5 0.07]. An analogousANOVA performed on the proportion of ‘‘know’’ responsesalso showed a significant condition effect [F(2, 20) 5 7.78,P < 0.01] and a trend toward significance for the group 3condition interaction [F(2, 20) 5 2.96, P 5 0.07]. In thiscase, however, read and anagrammed words received more‘‘know’’ responses than unstudied words [F(1, 10) 5 8.86 and14.34, P < 0.01 in both cases] but they did not differ fromeach other [F(1, 10) 5 1.46, P 5 n.s]. Moreover, while in theamnesic group the anagrammed words received more ‘‘know’’responses than read [F(1, 10) 5 4.37, P 5 n.s] and unstudiedwords [F(1, 10) 5 9.28, P < 0.01], in the group of healthysubjects the rate of ‘‘know’’ responses was higher for the read andanagrammed than for the unstudied words [F(1, 10) 5 14.50and 5.08, P < 0.01 and P < 0.05, respectively], but they didnot differ reciprocally [F(1, 10) 5 0.74, P 5 n.s.].

The dual-process signal-detection model was used to incor-porate the response bias in the –R/K procedure (Yonelinas andJacoby, 1995). Since the subjects were instructed to respond‘‘remember’’ when an item was recollected, the probability of a‘‘remember’’ response was used as an estimate of recollection(i.e., R 5 remember). Thus, true recollection is estimated bysubtracting the proportion of false ‘‘remember’’ responses fromthe proportion of true ‘‘remember’’ responses and then dividingby the opportunity to observe a true ‘‘remember’’ response,R 5 (Rold 2 Rnew)/(1 2 Rnew). As the subjects were instructedto respond ‘‘know’’ when an item was familiar in the absenceof recollection [i.e., know 5 F (1 2 R)], familiarity was esti-mated as the probability of a ‘‘know’’ response given the itemwas not recollected [i.e., F 5 ‘‘know’’/(1 2 R)]. Then, the fa-miliarity estimate was converted to d 0.

For recollection (Fig. 2) an ANOVA with group (amnesicsvs. controls) as between-subject factor and condition (read vs.anagrams) as within-subject factor revealed nonsignificanteffects of group [F(1, 10) 5 1.29, P 5 n.s.] and condition[F(1, 10) 5 3.16, P 5 n.s.] but a significant group 3 condi-tion interaction [F(1, 10) 5 5.35, P 5 0.04]. Planned compar-ison revealed that amnesics and controls did not differ in theread words condition [F(1, 10) 5 0.02, P 5 n.s.], whereas atrend toward significance emerged for the reduced recollectionexhibited by amnesics compared to normal controls in the ana-gram condition [F(1, 10) 5 3.84, P 5 0.07]. Moreover, theamnesics’ performance was similar in the read and anagramconditions [F(1, 10) 5 0.11, P 5 n.s.], whereas the controlsexhibited significantly more recollection for anagrammed thanfor read words [F(1, 10) 5 12.55, P < 0.005].

An analogous ANOVA, conducted on the familiarity scores(Fig. 2) revealed a significant condition main effect [F(1, 10) 55.12, P < 0.05] due to overall higher scores for words that had beengenerated than for words that had been read. By contrast, neitherthe group effect [F(1, 10)5 1.56, P5 n.s.] nor the group3 condi-tion interaction [F(1, 10) 5 2.75, P 5 n.s.] was significant.These results are indicative of the fact that the two groups didnot differ for familiarity in either read or anagrammed words.

Also in this case, the comparable familiarity scores in thetwo groups of patients were unlikely due to a floor effect inthe score distribution of healthy controls. Indeed, in no casedid the score distribution relative to the recollection and famili-arity estimates for both read and anagrammed words in thegroup of healthy controls significantly depart from normality(Kolmogorov-Smirnov d 0 ranged from 0.15 to 0.18). More-over, the number of healthy subjects who scored more than 1SD below the mean of their group did not differ for estimatesof recollection and familiarity.

Table 6 reports the estimates of recollection and familiarityderived from performance on the R/K paradigm by eachpatient in the amnesic group. Also in this case, the relativelyhigh variation coefficients in the score distribution of the groupof healthy controls (ranging from 0.57 to 1.0) prevented a for-mal statistical analysis of the individual patients’ performances.However, a qualitative analysis of the data confirmed thatamnesic patients made particularly poor estimates of recollec-tion for anagrammed words. Indeed, while one patient (CM)obtained a recollection estimate below the average of normalcontrols for the anagrammed words and another patientobtained below average recollection estimates for both read andanagrammed words, only one patient (AR) obtained a belowaverage familiarity estimate for the read words.

DISCUSSION

The present study was aimed at assessing the relative contri-bution of recollection and familiarity processes to the recogni-tion performance of a group of patients with severe amnesiaresulting from selective hippocampal damage. A critical issue,which deserves preliminary discussion, is the accurate localiza-tion of the brain damage in the patients with amnesia. Indeed,in order to derive reliable conclusions from our data about thespecific role of the hippocampus in the recognition processes,we had to be reasonably confident that the damage in ourpatients selectively involved the hippocampal formation whileleaving substantially unaffected the surrounding parahippocam-pal gyrus and, in particular, the perirhinal cortex, whose critical

TABLE 6.

Estimates of Recollection and Familiarity for Each Amnesic Patient

Derived From the Performance on the Remember–Know Procedure

Recollection (probability) Familiarity (d0)

Read Anagrammed Read Anagrammed

TE 0.22 (11.11) 0.34 (20.43) 0.37 (20.88) 0.92 (20.36)

CM 0.20 (10.02) 0.14 (21.19) 0.53 (20.68) 0.57 (20.90)

AR 0.00 (20.93) 0.08 (21.40) 0.00 (21.33) 1.89 (11.12)

CC 0.49 (11.40) 0.24 (20.82) 0.51 (20.71) 0.77 (20.59)

The number of standard deviations above or below the mean of healthy subjectsis given in parentheses.

476 TURRIZIANI ET AL.

Hippocampus

role in familiarity processes has been repeatedly emphasized(e.g., Bowles et al., 2007; Skinner and Fernandes, 2007). Inthis regard, in view of the hypoxic-induced pathology in 3 outof 4 of our patients, it should be noted that the neurobiologicalsubstrate of the cognitive deficits resulting from hypoxic eventsis still an unresolved issue. Indeed, one comprehensive reviewconcluded that in humans the cerebral cortex and basal gangliaare more frequently damaged than the hippocampus and thatselective hippocampal damage is actually an infrequent out-come of cerebral hypoxia (Caine and Watson, 2000). Whileacknowledging the difficulty of ruling out the possible presenceof extrahippocampal pathology simply based on the results of aneuroradiological exam, the manual segmentation of MTLstructures clearly demonstrated significant atrophy of the hip-pocampi and preserved volume of the perirhinal cortex in allthe participating amnesic patients. Only one patient, CM,showed atrophic changes at the level of some portion of theparahippocampal gyrus, but also in this case there was no evi-dence of a reduction in the volume of the perirhinal cortex.Noteworthy, the MR images of the three hypoxic patientsreported in the present study (together with those of two otherpatients) were also submitted to a semiautomatized voxel-basedmorphometry analysis that confirmed the selectivity of the hip-pocampal atrophy in the face of substantially preserved para-hippocampal structures in the group as a whole and in eachindividual subject (Di Paola et al., submitted for publication).

A very interesting outcome of the present study is that essen-tially the same pattern of results emerged from the use of botha subjective-based (R/K) and an objective-based (PDP) methodfor estimating the contribution of recollection and familiarityto the recognition performance. Indeed, consistent with a richliterature on the effects of generation on recollection processes(Jacoby, 1993; e.g., Gardiner, 1988), in both PDP and R/Kprocedures the healthy controls disclosed significantly higherrecollection estimates for words that had been anagrammedthan for words that had been read. Instead, patients with am-nesia obtained recollection scores that did not differ for wordsthat had been generated or that had been read. As a conse-quence, the recollection estimates for words that had been gen-erated were significantly reduced in the group of amnesics ascompared to the healthy controls while no difference wasdetected for words that had been simply read during the studyphase. A consistent pattern of results in the R/K and PDPparadigms also emerged for the estimates of familiarity. Here,in fact, both healthy controls and amnesic patients recognizedas familiar more words that had been generated than wordsthat had been read, and in no case did a significant differenceemerge between the two groups. It should be noted that essen-tially the same pattern of results emerged in the amnesicpatients from both the group comparison and the analysis ofindividual patients’ performances. In fact, each of the amnesicpatients obtained below average recollection estimates for theanagrammed words in at least one of the two experimentalparadigms; however, their recollection estimates for read wordsand familiarity estimates for read and anagrammed words werein the range of healthy controls. Interestingly, such a qualitative

pattern of performance was also observed in the only patient inour group (CM) who presented, besides hippocampal atrophy,a size reduction of some of the parahippocampal cortices (rightentorhinal and left parahippocampal). In view of currenthypotheses emphasizing the critical role of the perirhinal cortexin familiarity processes (Bowles et al., 2007; Skinner and Fer-nandes, 2007), the preserved familiarity in CM is likely due tothe substantial sparing of this portion of the parahippocampalgyrus.

To summarize, the results of the present study support thehypothesis of a specific role of the hippocampus in recollectionprocesses. Indeed, a group of patients with atrophic changesconfined to the hippocampal formation revealed a normal con-tribution of familiarity but a reduced contribution of recollec-tion to their recognition memory performance. The conver-gence of results obtained using the R/K and PDP paradigmssupports the robustness and authenticity of the dissociationbetween normal familiarity and impaired recollection processesin our group of patients. The finding that poor recollection inpatients with amnesia was observed for words that had beengenerated but not for words that had been read is consistentwith previous findings from a mixed group (hippocampal andextra-hippocampal) of patients with amnesia (Verfaellie andTreadwell, 1993) and is likely due to the particularly low recol-lection estimates exhibited by NCs for the words that had beenread; this, in turn, may be due to the fact that when readingand generating words are intermixed in a single study phase (asin the present study), the words that are read are subject to avery shallow encoding that interferes with their subsequent rec-ollection (Verfaellie and Treadwell, 1993).

The results of the present study add to a growing literatureon the effects of brain damage on the recollection and familiar-ity components of episodic memory (see, for recent reviews,Aggleton et al., 2005; Cipolotti and Bird, 2006; Skinner andFernandes, 2007). As noted earlier, the results of neuropsycho-logical investigations on patients with damage confined to theMTL region are critical for the debate between the sustainersof recollection and familiarity as distinct memory processesunderlain by independent neural circuits (hippocampus andparahippocampal cortices, respectively) and those who claimthat recollection and familiarity are different manifestations ofa unitary memory system, the former being an expression of‘‘strong’’ memory traces, rich with intercorrelated contextualdetails, and the latter, which rises from ‘‘weak’’ memory traces,lacking in associative information. In fact, our findings of poorrecollection but normal familiarity contribution to recognitionperformance in amnesic patients with damage restricted to thehippocampus is supportive of the dual-process models, suggest-ing a specific role of the hippocampus in recollection but notin familiarity processes. Similar findings have been reported ina number of single case investigations of hippocampal amnesicpatients (Baddeley et al., 2001; Holdstock et al., 2002; Bastinet al., 2004; Aggleton et al., 2005) and in a small group ofpatients with mild memory disorder because of hypoxia (Yone-linas et al., 2002). However, other authors failed to detect thedissociation we observed in our group of patients. In particular,

RECOLLECTION AND FAMILIARITY IN THE HIPPOCAMPUS 477

Hippocampus

Manns et al. (2003) and Wais et al. (2006) reported a compa-rable decrement in the recollection and familiarity componentsof recognition in groups of patients with hippocampal amnesia.The reason for these controversial results is unclear and likelyinvolves procedural differences in the estimate of recollectionand familiarity and discrepancies in the samples of amnesicpatients recruited. As noted earlier, one possible explanation ofthe discrepant findings is the use of subjective-based or objec-tive-based measurement methods for estimating recollectionand familiarity. Indeed, even though in the present study con-vergent results were obtained using subjective-based (R/K) andobjective-based (PDP) measurement paradigms, we could notexclude that inconsistencies in the previous literature were due,among other things, to the use of subjective methods of meas-urements in a population of patients, such as amnesics, thathave difficulty in reliably introspecting their own memoryperformances. As a matter of fact, both of the two above-mentioned studies reporting impaired familiarity in amnesicpatients were based on subjective-based measurement methods(R/K and ROC).

Another possible reason for discrepant results in the previousliterature is the variable severity of the memory disorder in thepatients participating in the different studies. Indeed, accordingto a theoretical position that assumes a role for the hippocam-pus in both recollection and familiarity processes (Squire,2004) and to the extent to which the hippocampal damagewould deteriorate the more complex recollective process first,and only later the more rudimentary familiarity process, it isexpected that less severe amnesic patients will exhibit a dissocia-tion between impaired recollection and preserved familiaritywhile more severe amnesic patients will be impaired in bothkinds of processes. Consistent with this view, in the previousgroup studies that investigated recollection and familiarity inpatients with selective hippocampal damage, the discrepantresults obtained may have been related to the different severityof the memory disorder in the two groups of patients. In fact,the patients in Manns et al. (2003) and Wais et al. (2006)studies, who were equally impaired in recollection and familiar-ity, had a more severe memory deficit than the patients inYonelinas et al. study (2002), whose familiarity was preserved.Our results do not agree with this conclusion. Indeed, thememory deficit in our group of patients was, at the very least,probably as severe as in Manns et al. (2003) and Wais et al.(2006) samples, and they obtained normal familiarity scores.Instead, our results demonstrate that to the extent to which theparenchymal damage is confined to the hippocampus, amnesicpatients present a normal contribution of familiarity and areduced contribution of recollection to their recognition per-formances, irrespective of the overall severity of their memorydisorder.

As a final tentative explanation for the discrepancy betweencurrent and previously reported results, the possibility shouldbe considered that the amnesic patients who participated in ourstudy and in previous ones differed as to the extension of braindamage to extra-MTL cerebral regions. In particular, in thelight of recent evidence of poor familiarity in patients with Par-

kinson’s disease (Davidson et al., 2006), and since basal gangliaare among the most consistently damaged structures followingcerebral hypoxia (Caine and Watson, 2000), one wonderswhether the familiarity deficit in patients with hippocampalamnesia might actually reflect extension of the hypoxic damageto the basal ganglia. In fact, morphological and volumetricanalysis of basal ganglia was not reported in previous studies.As for the presently reported amnesic patients, accurate exami-nation of cerebral magnetic resonance imaging (MRI) did notshow any atrophic or focal change at this level.

The finding of poor recollection but normal familiarity proc-esses in patients with damage confined to the hippocampus issuggestive but not conclusive evidence for the anatomical inde-pendence, in the context of MTL, of neural circuits supportingrecollection and familiarity processes. In fact, a stronger confir-mation of this hypothesis would come from evidence of a dou-ble neuropsychological dissociation between hippocampalpatients with selective impairment of recollection, but sparedfamiliarity, and patients in whom the anatomical damageinvolved the parahippocampal structures (particularly, the peri-rhinal cortex), leaving unaffected the hippocampus, and whoshowed the opposite pattern of impairment (i.e., normal recol-lection but poor familiarity). In fact, while we have on recordonly patients with poor recollection but preserved familiarity,the hypothesis that the two memory processes are not manifes-tations of reciprocally independent memory mechanisms butrather the expression of the access to memory traces of differentstrength (Wixted and Squire, 2004; Slotnick and Dodson,2005) cannot be dismissed. Unfortunately, there is scarce evi-dence for such a double dissociation in the literature. However,a very recent report by Bowles et al. (2007) described the caseof a patient who underwent surgical resection of left anteriortemporal-lobe structures for treatment of intractable epilepsy.The resection included a large portion of the perirhinal cortexbut spared the hippocampus. The results of four experimentsindicated that this patient exhibited impaired familiarity withpreserved recollection.

In addition to neuropsychological data, other sources ofevidence support the independence of the neural circuits under-lying recollection and familiarity. In fact, an event-relatedpotentials study identified two functionally and temporallydissociable anatomical populations, respectively, supportingrecollection and familiarity processes (Rugg and Yonelinas,2003). Moreover, a growing number of fMRI studies havereported dissociable patterns of activity for measures of famili-arity and recollection (Davachi et al., 2003; Ranganath et al.,2004; Weis et al., 2004; Eldridge et al., 2005; Henson et al.,2005; Yonelinas et al., 2005; Daselaar et al., 2006; Montaldiet al., 2006; see, for recent reviews, Skinner and Fernandes,2007). Some of these dissociations reflect the nature of theresponse—for example, activation for recollection and deactiva-tion for familiarity (Weis et al., 2004; Henson et al., 2005;Montaldi et al., 2006), while some others are structural. Forexample, using the R/K procedure, Yonelinas et al. (2005) andEldridge et al. (2005) found increased hippocampal activitycorrelated with reports of recollection but not familiarity with

478 TURRIZIANI ET AL.

Hippocampus

the learning episode. Variants of the PDP have linked hippo-campal activity to recollection and anterior parahippocampalactivity to familiarity (Davachi et al., 2003; Ranganath et al.,2004; Weis et al., 2004; Daselaar et al., 2006; Montaldi et al.,2006). Finally, use of confidence judgment revealed a doubledissociation between hippocampal and anterior parahippocam-pal cortex activity, the former associated with recollection andthe latter with familiarity (Daselaar et al., 2006).

In conclusion, we have provided evidence that recollectionbut not familiarity processes are impaired in the recognitionperformance of amnesic patients with pure hippocampal dam-age. The results of this group study on patients with severe am-nesia are in keeping with the results of previous single casereports of hippocampal amnesics (Baddeley et al., 2001; Hold-stock et al., 2002; Bastin et al., 2004; Aggleton et al., 2005)and with those of a previous group study on patients with amild memory disorder due to hypoxia (Yonelinas et al., 2002).Some new data from the literature describe the inverse dissocia-tion (i.e., poor familiarity but normal recollection) in patientswith Parkinson’s disease (Davidson et al., 2006) and, moreinterestingly, in patients with selective perirhinal damage(Bowles et al., 2007). The convergence of these neuropsycho-logical results with those deriving from functional neuroimag-ing investigation of healthy subjects strongly supports the viewthat recollection and familiarity are distinct cognitive processessupported by the activity of different neuronal populations. Inthe context of the MTL, the hippocampus proper is responsiblefor recollection, while neocortical associative areas in the para-hippocampal gyrus are likely involved in familiarity.

REFERENCES

Aggleton JP, Brown M. 1999. Episodic memory, amnesia, and the hip-pocampal-anterior thalamic axis. Behav Brain Sci 22:425–489.

Aggleton JP, Vann SD, Denby C, Dix S, Mayes AR, Roberts N, Yone-linas AP. 2005. Sparing of the familiarity component of recognitionmemory in a patient with hippocampal pathology. Neuropsycholo-gia 43:1810–1823.

Baddeley AD, Vargha-Khadem F, Mishkin M. 2001. Preserved recog-nition in a case of developmental amnesia: Implications for the ac-quisition of semantic memory? J Cogn Neurosci 13:357–369.

Bastin C, Linden M, Charnallet A, Denby C, Montaldi D, Roberts N,Andrew M. 2004. Dissociation between recall and recognition mem-ory performance in an amnesic patient with hippocampal damagefollowing carbon monoxide poisoning. Neurocase 10:330–344.

Bortolini V, Tagliavini C, Zampilli A. 1971. Lessico di frequenze dellalingua italiana contemporanea [Frequency of words in the contem-porary Italian language]. Milan, Italy: Garzanti.

Bowles B, Crupi C, Mirsattari SM, Pigott SE, Parrent AG, PruessnerJC, Yonelinas AP, Kohler S. 2007. Impaired familiarity with pre-served recollection after anterior temporal-lobe resection that sparesthe hippocampus. Proc Natl Acad Sci USA 104:16382–16387.

Brown MW, Aggleton JP. 2001. Recognition memory: What are theroles of the perirhinal cortex and hippocampus? Nat Rev Neurosci2:51–61.

Caine D, Watson JDG. 2000. Neuropsychological and neuropatholog-ical sequelae of cerebral anoxia: A critical review. J Int Neuropsy-chol Soc 6:86–99.

Carlesimo GA, Caltagirone C, Gainotti G. 1996. The mental deterio-ration battery: Normative data, diagnostic reliability and qualitativeanalyses of cognitive impairment. Eur Neurol 36:378–384.

Carlesimo GA, Buccione I, Fadda L, Graceffa A, Mauri M, Lo RussoS, Bevilacqua G, Caltagirone C. 2002. Standardizzazione di duetest di memoria per uso clinico: Breve racconto e Figura di Rey.Nuova Rivista di Neurologia 12:3–13.

Cipolotti L, Bird CM. 2006. Amnesia and the hippocampus. CurrOpin Neurol 19:593–598.

Cipolotti L, Bird C, Good T, Macmanus D, Rudge P, Shallice T.2006. Recollection and familiarity in dense hippocampal amnesia:A case study. Neuropsychologia 44:489–506.

Collins DL, Neelin P, Peters TM, Evans AC. 1994. Automatic 3Dintersubject registration of MR volumetric data in standardizedTalairach space. J Comput Assist Tomogr 18:192–205.

Daselaar SM, Fleck MS, Prince SE, Cabeza R. 2006. Triple dissocia-tion in the medial temporal lobes: Recollection, familiarity, andnovelty. J Neurophysiol 96:1902–1911.

Davachi L, Mitchell JP, Wagner AD. 2003. Multiple routes to mem-ory: Distinct medial temporal lobe processes build item and sourcememories. Proc Natl Acad Sci USA 100:2157–2162.

Davidson PSR, Anaki D, Saint-Cyr JA, Chow TW, Moscovitch M. 2006.Exploring the recognition memory deficit in Parkinson’s disease: Esti-mates of recollection versus familiarity. Brain 129:1768–1779.

Eldridge LL, Engel SA, Zeineh MM, Bookheimer SY, Knowlton BJ.2005. A dissociation of encoding and retrieval processes in thehuman hippocampus. J Neurosci 25:3280–3286.

Fortin NJ, Wright SP, Eichenbaum H. 2004. Recollection-like mem-ory retrieval in rats is dependent on the hippocampus. Nature431:188–191.

Gardiner JM. 1988. Functional aspects of recollective experience.Mem Cogn 16:309–313.

Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frack-owiak RS. 2001. A voxel-based morphometric study of ageing in465 normal adult human brains. Neuroimage 14:21–36.

Henson RN, Hornberger M, Rugg MD. 2005. Further dissociatingthe processes involved in recognition memory: An FMRI study.J Cogn Neurosci 17:1058–1073.

Holdstock JS, Mayes AR, Roberts N, Cezayirli E, Isaac CL, O’ReillyRC, Norman KA. 2002. Under what conditions is recognitionspared relative to recall after selective hippocampal damage inhumans. Hippocampus 12:341–351.

Insausti R, Juottonen K, Soininen H, Insausti AM, Partanen K, VainioP, Laakso MP, Pitkanen A. 1998. MR volumetric analysis of thehuman entorhinal, perirhinal, and temporopolar cortices. Am JNeuroradiol 19:659–671.

Jacoby LL. 1991. A process dissociation framework: Separatingautomatic from intentional uses of memory. J Mem Lang 30:513–541.

Jacoby LL, Toth JP, Yonelinas AP. 1993. Separating conscious andunconscious influences of memory: Measuring recollection. J ExpPsychol Gen 123:216–219.

Mandler G. 1980. Recognizing: The judgement of previous occur-rence. Psychol Rev 87:252–271.

Manns JR, Hopkins RO, Reed JM, Kitchener EG, Squire LR. 2003.Recognition memory and the human hippocampus. Neuron37:171–180.

Montaldi D, Spencer TJ, Roberts N, Mayes AR. 2006. The neuralsystem that mediates familiarity memory. Hippocampus 16:504–520.

Nelson HE. 1976. A modified card sorting test sensitive to frontallobe defects. Cortex 12:313–324.

Nocentini U, Di Vincenzo S, Panella M, Pasqualetti P, Caltagirone C.2002. La valutazione delle funzioni esecutive nella pratica neuropsi-cologica:dal Modified Card sorting Test al Modified Card SortingTest-Roma Version. Dati di standardizzazione. Nuova Rivista diNeurologia 12:14–24.

RECOLLECTION AND FAMILIARITY IN THE HIPPOCAMPUS 479

Hippocampus

Orsini A, Laicardi C. 1997. WAIS-R. Contributo alla taratura italiana.Firenze: Organizzazioni Speciali.

Pruessner JC, Li LM, Serles W, Pruessner M, Collins DL, Kabani N,Lupien S, Evans AC. 2000. Volumetry of hippocampus and amyg-dale with high-resolution MRI, three-dimensional analysis software:Minimizing the discrepancies between laboratories. Cerebr Cortex10:433–442.

Pruessner JC, Kohler S, Crane J, Pruessner M, Lord C, Byrne A,Kabani N, Collins DL, Evans AC. 2002. Volumetry of temporopo-lar, perirhinal, entorhinal and parahippocampal cortex from highresolution MR images: Considering the variability of the collateralsulcus. Cerebr Cortex 12:1342–1353.

Ranganath C, Yonelinas AP, Cohen MX, Dy CJ, Tom SM, D’EspositoM. 2004. Dissociable correlates of recollection and familiaritywithin the medial temporal lobes. Neuropsychologia 42:2–13.

Raven JC. 1949. Progressive matrices (1947), sets A, Ab, B: board andbook forms. London: Lewis.

Rey A. 1941. L’examon psychologique dans les cas d’ence phalopathietraumatique. Arch de Psychol 28:286–340.

Rey A. 1958. Memorisation d’une serie de 15 mots en 5 repetitions.In: Rey A, editor. L’examen clinique en psycologie. Paris: PressesUniversitaires de France.

Rugg MD, Yonelinas AP. 2003. Human recognition memory: A cogni-tive neuroscience perspective. Trends Cogn Sci 7:313–319.

Skinner EI, Fernandes MA. 2007. Neural correlates of recollection andfamiliarity: A review of neuroimaging and patient data. Neuropsy-chologia 45:2163–2179.

Slamecka NJ, Graf P. 1978. The generation effect: Delineation of aphenomenon. J Exp Psychol: Hum Learn Mem 4:592–604.

Slotnick SD, Dodson CS. 2005. Support for a continuous (single-pro-cess) model of recognition memory and source memory. MemCogn 33:151–170.

Spinnler H, Tognoni G. 1987. Standardizzazione e taratura italiana ditests neuropsicologici. Ital J Neurol Sci 8:35–38.

Squire LR. 2004. Memory systems of the brain: A brief history andcurrent perspective. Neurobiol Learn Mem 82:171–177.

Tulving E. 1985. Memory and consciousness. Can Psychol 26:1–12.Verfaellie M, Treadwell JR. 1993. Status of recognition memory in

amnesia. Neuropsychology 7:5–13.Wais PE, Wixted JT, Hopkins RO, Squire LR. 2006. The hippocam-

pus supports both the recollection and the familiarity componentsof recognition memory. Neuron 49:459–466.

Wechsler D. 1981. WAIS-R, Wechsler Adult Intelligence Scale Re-vised. New York: The Psychological Corporation.

Weis S, Specht K, Klaver P, Tendolkar I, Willmes K, Ruhlmann J,Elger CE, Fernandez G. 2004. Process dissociation between con-textual retrieval and item recognition. Neuroreport 15:2729–2733.

Wixted JT, Squire LR. 2004. Recall and recognition are equallyimpaired in patients with selective hippocampal damage. CognAffect Behav Neurosci 4:58–66.

Yonelinas AP. 2002. The nature of recollection and familiarity: Areview of 30 years of research. J Mem Lang 46:441–517.

Yonelinas AP, Jacoby LL. 1995. The relation between rememberingand knowing as bases for recognition: Effects of size congruency.J Mem Lang 34:622–643.

Yonelinas AP, Kroll NEA, Dobbins I, Lazzara MM, Knight RT. 1998.Recollection and familiarity deficits in amnesia: Convergence ofremember-know, process dissociation and receiver operating charac-teristic data. Neuropsychology 12:323–339.

Yonelinas AP, Kroll NEA, Quamme JR, LazzaraMM, SauveM-J,WidamanKF, Knight RT. 2002. Effects of extensive temporal lobe damage ormild hypoxia on recollection and familiarity. Nat Neurosci 5:1236–1241.

Yonelinas AP, Otten LJ, Shaw KN, Rugg MD. 2005. Separating thebrain regions involved in recollection and familiarity in recognitionmemory. J Neurosci 25:3002–3008.

480 TURRIZIANI ET AL.

Hippocampus