Frequency judgements of abstract designs by patients with diencephalic, temporal lobe or frontal lobe lesions

\

Pergamon

PII] S9917Ð2821"87#99939Ð1

Neuropsycholo`ia\ Vol[ 25\ No[ 01\ pp[ 0276Ð0285\ 0887 Þ 0887 Elsevier Science Ltd[ All rights reserved Printed in Great Britain 9917Ð2821:87 ,08[99¦9[99

Frequency judgements of abstract designs by patients with diencephalic\ temporal lobe or frontal lobe lesions NICOLA STANHOPE\% ELIZABETH GUINAN$ and MICHAEL D[ KOPELMAN$  Psychology Department\ Institute of Psychiatry\ De Crespigny Park\ Denmark Hill\ London SE4 7AF\ U[K[ $ Neuropsychiatry and Memory Disorders Clinic\ Division of Psychiatry and Psychology\ UMDS\ St Thomas| Hospital\ Lambeth Palace Road\ London SE0 6EH\ U[K[ "Received 3 February 0886^ accepted 08 February 0887#

Abstract*Patients with diencephalic\ temporal lobe or frontal lobe lesions were compared with healthy controls on a frequency judgement task[ The three patient groups were disproportionately impaired at estimating how often a series of abstract designs had been presented relative to controls[ Diencephalic and temporal lobe patients did not di}er from each other[ It is argued that the results may re~ect a {core| memory de_cit in the temporal lobe patients[ The impairment in the frontal patients may re~ect their di.culty in making an organised search in memory for multiple traces of an item\ while the de_cit shown by the diencephalic patients "particularly those with Korsako} syndrome# may be due to the combined e}ects of a generally poor memory and superimposed frontal pathology[ Þ 0887 Elsevier Science Ltd[ All rights reserved[ Key Words] context memory^ frequency estimation^ amnesic patients

elman ð01Ł found only very weak correlations between temporal context memory and frontal dysfunction in Korsako} and Alzheimer patients\ and none with a CT scan measure of frontal atrophy\ although there were signi_cant correlations with the severity of the core mem! ory disorder[ In contrast\ Sagar et al[ ð15Ł found that patient H[M[\ who is amnesic following bilateral tem! poral lobe excision ð16Ł\ showed relatively preserved recency and frequency discriminations under some cir! cumstances despite severely impaired content recog! nition[ By neuropsychological convention\ frequency judge! ments "i[e[\ estimating the number of times events occur# have often been studied alongside recency judgements ð8\ 07Ł[ As a consequence\ frequency has come to be regarded as an aspect of context memory\ which\ like other aspects of context\ might depend on the integrity of the frontal lobes[ Whether frequency memory should really be seen as a form of context memory is an empirical question which is considered later in this paper[ In their study of frequency judgement ability\ Smith and Milner ð21\ 22Ł presented a series of abstract designs to patients with unilateral frontal damage and to patients who had had temporal lobe excisions[ The results indi! cated that defective frequency estimation was not sec!

Introduction A growing body of research suggests that impairments in some kinds of context memory can be attributed to fron! tal lobe pathology[ Frontal lobe lesions produce de_cits in memory for temporal context ð08Ł\ temporal sequ! encing ð29Ł\ in memory for the source of information ð09\ 12\ 17Ł and spatial memory ð20Ł[ The putative role of the frontal lobes in context memory was given some apparent support by Squire ð23Ł who reported an association between impaired temporal context "recency# memory and frontal dysfunction in Korsako} patients "with pre! sumed diencephalic and frontal lobe pathology# but not in temporal lobe patients[ Other studies suggest that lesions in the diencephalon or temporal lobes may also lead to disproportionate impairments in some aspects of context memory[ The evidence is mixed\ and tasks have varied somewhat from study to study[ Parkin and Hunkin ð11Ł reported that a patient with no frontal pathology was disproportionately impaired on tests of temporal context memory[ Kop!

% Corresponding ¦9060!697!2386

author[

Tel[]

¦9060!808!2113^

Fax]

0276

0277

N[ Stanhope et al[:Frequency judgements of abstract designs

ondary to a de_cit in recognition] frontal\ but not temporal lobe patients exhibited normal recognition\ whereas only frontal patients\ particularly those with right!sided lesions\ were impaired in making frequency judgements[ Huppert and Piercy ð8Ł showed that Kor! sako} patients were impaired in making frequency "and recency# judgements\ and\ using the same paradigm\ Mayes et al[ ð06Ł found similar de_cits in other groups of amnesic patients[ Meudell et al[ ð07Ł found that when Korsako} patients| picture recognition was matched with controls\ frequency judgements were still dis! proportionately impaired\ while Strauss et al[ ð24Ł reported a disproportionate de_cit in frequency dis! crimination in Korsako} patients relative to target recall\ although this was also impaired[ These studies do not\ however\ allow us to determine how far Korsako} pat! ients| frequency memory impairments may be associated with frontal lobe dysfunction[ It is not disputed that frontal lobe lesions can result in de_cits in making frequency judgements[ What is less clear is whether such de_cits in patients with lesions else! where are attributable to concomitant frontal lobe dys! function or whether they are associated more with target memory performance and so re~ect a more general mem! ory impairment[ There is a lack of studies which have examined the ability to make frequency judgements in patients with focal lesions con_ned largely to either the diencephalon\ the temporal lobes or the frontal lobes^ most investigate only one\ or at most\ two of these patient groups at a time[ In addition\ in many cases\ the presumed lesion site has usually been made on the basis of clinical diagnosis rather than by concurrent neuroradiological evidence[ The lesion site in our patients was con_rmed by magnetic resonance imaging "MRI# and positron emission tom! ography "PET# "using labelled ~uoro!deoxyglucose# scans^ the _ndings from these scans are brie~y sum! marised below[ The present study investigates memory for item frequency in these patient groups and employs a design based on Smith and Milner|s ð22Ł work[

Method Subjects Three groups of patients were selected on the basis of having predominantly focal lesions producing memory impairment as determined by clinical observation\ CT scan and\ where relevant\ EEG data[ Fifteen patients had diencephalic lesions\ 03 had temporal lobe lesions and 04 had frontal lobe lesions[ Twenty healthy controls matched for age and full!scale IQ also took part[ The diencephalic group consisted of thirteen patients with alcoholic Korsako} syndrome[ Twelve of these had either a documented history or residual signs of a Wernicke episode which preceded their amnesic syndrome[ On the basis of clinical evidence and an IQ!General MQ "Memory Quotient# dis! crepancy of at least 19 points\ all 02 Korsako} patients had

memory impairments disproportionate to any other cognitive de_cit ð02\ 26Ł[ The two other patients in the diencephalic group had memory de_cits following surgical excision and radiotherapy for treat! ment of a pituitary adenoma[ One of these patients is described in greater detail elsewhere ð4Ł[ She has an IQ!General MQ discrepancy of 38 points\ whilst the other has an IQ!General MQ discrepancy of 16 points[ Plotting the radiotherapy plan! ning _elds against these two patients| MRI scans indicates that the structures a}ected by ×89) of the irradiation included the anterior thalamus\ the mammillo!thalamic tracts and the fornix ð4Ł[ Therefore\ these two patients were treated as cases of dien! cephalic amnesia[ The temporal lobe group included nine patients with a prob! able or de_nite diagnosis "antibody con_rmed# of herpes sim! plex encephalitis[ Four other patients were amnesic as a result of hypoxic episodes[ One patient had hypoxic brain damage following a heroin overdose which resulted in prolonged uncon! sciousness^ his CT scan showed enlarged temporal horns of the ventricles bilaterally\ indicating temporal lobe atrophy[ Ano! ther patient had attempted suicide by hanging^ his CT scan showed a large area of infarction in the left temporal lobe and his MRI scan con_rmed left temporal atrophy[ The CT scan of the third patient showed evidence of left temporo!parietal infarction following a respiratory arrest[ Little was seen on the CT scan of the fourth patient\ but visual inspection of his PET scan suggested focal hypometabolism in the temporal lobes[ The _nal patient in the temporal lobe group had a prolonged history of complex partial seizures with bilateral temporal lobe foci evident on EEG and his CT scan also showed some tem! poral lobe atrophy[ He had been referred clinically because of a moderate degree of memory impairment\ but there was no evidence of generalised dementia at either his initial assessment or at follow up three years later[ The frontal lobe group consisted of nine patients who had had stereotactic subcaudate tractotomy for treatment of intractable a}ective disorders[ In this operation\ yttrium rods are implanted in the frontal lobes and a surrounding area of radiation necrosis\ as well as a much larger area of oedema results "although the oedema always seems to spare a margin of cortical tissue#[ The patients were tested one to two weeks after the operation "or\ if there was prolonged confusion\ in the third post!operative week# because in a longitudinal study\ Kartsounis et al[ ð00Ł found that these patients behaved most like patients with large frontal lesions during this time[ Kartsounis et al[ ð00Ł found statistically signi_cant impairments on tests of executive:frontal lobe function and recognition memory two weeks post!oper! atively although these were not present either pre!operatively "despite severe depression# or six months later[ Six other patients with focal frontal lesions were also tested[ One patient had had a bilateral frontal tractotomy in 0863\ 07 years before testing[ The other _ve patients had lesions resulting from a glioma\ haematoma\ infarcts\ or craniotomy[ Three of the six patients had right frontal lesions\ one had a left frontal lesion\ and two had bilateral lesions[ Three of these patients had lesions extending into the lateral frontal cortex^ the others had more medially situated lesions\ as did the tractotomy pati! ents\ whose yttrium rods were placed across the medial and mid!frontal regions[ Table 0 shows the mean age\ estimated premorbid IQ "NART!R#\ and current IQ "WAIS!R# across the four groups ð10\ 28Ł[ The groups did not di}er signi_cantly in mean age\ F"2\59#  0[44\ P  9[10\ premorbid IQ or current full!scale IQ "both Fs ³ 0#\ although the frontal lobe patients showed a mean NART!WAIS IQ discrepancy of about 8[4 points\ against 7[0 points in the temporal lobe group\ 6[9 in the diencephalic group\ and 1[1 points in the healthy controls[ Table 0 also shows the mean immediate "general# and delayed memory quotients "GMQ and DMQ respectively# in the patient

N[ Stanhope et al[:Frequency judgements of abstract designs

0278

Table 0[ Mean scores and standard deviations for the three patient groups and controls for age\ premorbid IQ "NART!R#\ current full!scale IQ "WAIS!R# and anterograde memory "WMS!R# Control

Age Age range NART!R IQ WAIS!R FSIQ General MQ Delayed MQ

Diencephalic

Temporal

Frontal

Mean

S[D[

Mean

S[D[

Mean

S[D[

Mean

S[D[

P

34[8 11Ð62 093[2 091[0 092[0 094[1

06[2

43[3 32Ð57 097[6 090[6 69[9 54[9

7[1

34[0 11Ð56 093[9 84[8 57[3 54[8

05[2

34[6 13Ð59 094[0 091[0 72[0 62[4

09[1

ns

01[8 05[2 06[1 04[3

ns ns 9[990 9[990

03[3 04[3 05[3 08[9

00[7 06[6 19[4 10[9

8[3 00[6 03[5 03[9

Signi_cance compared to controls is shown by underlining "NewmanÐKeuls test#[

Table 1[ Mean scores and standard deviations for the three patient groups and controls on tests of frontal lobe function Control

FAS verbal ~uency Card sorting] categories percent perseverations Cognitive estimates

Diencephalic

Temporal

Frontal

Mean

S[D[

Mean

S[D[

Mean

S[D[

Mean

S[D[

P

39[2

02[6

21[1

04[4

21[7

02[5

12[6

02[6

9[90

4[9 02[0 2[8

0[5 03[6 1[6

2[6 21[8 5[8

1[9 20[3 3[3

3[3 12[8 5[0

0[6 14[3 2[1

2[2 30[2 7[8

0[8 29[9 2[8

9[94 9[90 9[990

Signi_cance compared to controls is shown by underlining "NewmanÐKeuls test#[

groups ð2Ł[ The diencephalic and temporal lobe patients were severely and similarly impaired[ The frontal lobe patients showed a lesser but quite severe degree of memory impairment relative to healthy controls[ Taking GMQ and DMQ together\ there was an overall signi_cant main e}ect of group\ F"2\59#  08[09\ P ³ 9[9990\ and post hoc NewmanÐKeuls tests showed that all three patient groups were worse than controls "P ³ 9[94#\ but that the patient groups did not di}er sig! ni_cantly from each other[ Three tests of {frontal lobe| function\ the FAS verbal ~uency test ð1Ł\ the Cognitive Estimates test ð18Ł and the Modi_ed Wisconsin Card!Sorting test ð19Ł were also administered to all subjects[ The results are shown in Table 1[ There were sig! ni_cant group di}erences on all these tests and the frontal lobe patients showed the poorest mean scores across the four measures[ NewmanÐKeuls tests showed that the frontal patients were worse than controls on verbal ~uency\ percent per! severations "card!sorting# and cognitive estimates "P ³ 9[94# There were no other signi_cant di}erences[

Summary of quantitative MRI and PET _ndings For present purposes\ we will simply report _ndings pertinent to identifying focal lesions\ atrophy\ or hypometabolism within each patient group[ Detailed descriptions of the methods of imaging\ image analysis and statistical analyses will be reported elsewhere[ All 02 Korsako} patients showed evidence of thalamic atro! phy on MRI\ and as a group di}ered signi_cantly "P ³ 9[90# from controls[ Nine of the 02 patients also showed evidence of mammillary body atrophy[ In terms of global brain volume\ there was some degree of atrophy\ as would be expected[ However\ on Duncan|s Multiple Range test\ this was signi_cant at only the P  9[94 level\ and was mainly attributable to two

patients[ After excluding these two patients\ the Korsako} group did not di}er signi_cantly from the controls in terms of global atrophy[ However\ these two patients| performance in the present study was typical of the rest of the Korsako} group[ On PET region!of!interest analysis\ there was a trend to lower frontal metabolism in the Korsako} patients\ largely attribu! table to changes in four patients\ although these four did not di}er from the rest of the group on frontal:executive test per! formance[ Little has emerged from the MRI and PET analyses of the pituitary irradiation group to date\ and they did not di}er signi_cantly from the controls[ However\ an apparently normal MRI scan does not mean that the underlying tissue is normal ð03\ 27Ł[ The fact that the irradiation to the anterior thalamus\ fornix and mammillo!thalamic tract "see above# was followed by a clinically and neuropsychologically severe anterograde memory disorder\ in the absence of any global cognitive impair! ment\ would appear to con_rm that there must be damage to these structures\ albeit at a microscopic level[ The herpes encephalitis group showed temporal lobe abnor! malities on MRI and PET clearly apparent to the naked eye in all cases[ In eight out of nine patients\ these were asymmetrical\ but despite this\ there were signi_cant abnormalities in total temporal lobe volume and metabolism[ These included sig! ni_cant reductions in total hippocampal "P ³ 9[94#\ par! ahippocampal "P ³ 9[94# and combined hippocampal and parahippocampal volumes "P ³ 9[90# on MRI[ On PET region! of!interest analysis\ they showed statistically signi_cant hypo! metabolism in the medial\ inferior\ and anterior temporal regions\ but no signi_cant hypometabolism in any other brain regions[ As mentioned above\ three of the four anoxic patients had unequivocal CT scan evidence of temporal lobe changes when this was performed shortly after the cerebral insult[ The remain! ing patient appeared to have bilateral temporal lobe hypo!

0289

N[ Stanhope et al[:Frequency judgements of abstract designs

metabolism on initial visual inspection of his PET[ On MRI\ the anoxic group showed a signi_cant reduction in total hip! pocampal volume "P ³ 9[94#[ An unexpected _nding was that three of the four anoxic patients showed thalamic hypo! metabolism on PET[ Clinically this is not altogether surprising\ since anoxic patients sometimes show focal points of altered signal in subcortical grey matter[ However\ it was unexpected in view of the way these patients had been selected on clinical and CT evidence for inclusion in the study[ In brief\ there was evidence for both temporal lobe and thalamic changes in this group[ In the analyses below\ we have included these patients in the temporal lobe group\ i[e[\ in the manner in which they had been selected originally on the clinical and CT evidence\ but have also examined for di}erences in performance between the herpes encephalitis and anoxic groups[ The MRI and PET changes in the frontal tractotomy group were highly consistent from patient to patient[ The yttrium deposits were surrounded by an area of irradiation necrosis\ which\ in turn\ was engulfed by a large area of oedema[ This necrosis and oedema was accompanied by severe hypo! metabolism demonstrable on PET[ There was statistically sig! ni_cant hypometabolism in the dorso!lateral and orbito!medial frontal cortex\ in the anterior cingulate region\ as well as in the {whole frontal lobe|[ There was no statistically signi_cant hypometabolism in any other brain region and\ as expected\ no signi_cant reduction in total brain volume on MRI[ The focal frontal group had MRI!con_rmed frontal lesions "1 bilateral\ 2 right unilateral\ 0 left unilateral#[ On MRI\ this group did not di}er signi_cantly from controls in terms of total brain volume\ but they did di}er signi_cantly "P ³ 9[94# when the lesion volume in the frontal lobes was subtracted to give brain minus lesion volume[ In terms of PET\ pooling the data from the left and right hemispheres\ the focal frontal group showed statistically signi_cant hypometabolism in the dorso! lateral and orbito!medial frontal lobes\ anterior cingulate\ and {whole frontal lobes|[ There was no statistically signi_cant hypo! metabolism elsewhere[

Design and procedure The materials consisted of 21 black and white abstract designs generated on an Apple Macintosh MacDraw program[ Each design was presented on a 5×3 in[ white ~ash card[ For the study phase\ 05 of the designs were randomly assigned to one of the four frequency levels] 0\ 3\ 7\ and 01[ The total set comprised 52 items[ The number of designs at each frequency level varied so that seven designs appeared once\ _ve designs appeared four times\ three designs appeared eight times\ and one design appeared twelve times[ The repeated designs were systematically distributed throughout the set so that\ for exam! ple\ a design appearing four times would appear once in each quarter of the total set[ No two designs with the same frequency were presented in succession[ Although the particular fre! quencies and the number of items we used di}ered slightly from those in Smith and Milner|s study ð22Ł\ our design was essentially modelled on theirs[ Subjects were told they would be shown a series of cards\ each with an abstract design on it which they were to try to remember[ To ensure that subjects attended to the designs\ they were asked to say whether each consisted of straight lines only\ curved lines only\ or both straight and curved[ A card with the words\ {straight|\ {curved| and {both| was placed in front of the subject throughout the study phase[ They were also told that some of the designs would appear more than once\ but were not told why or that they would have to recall how many times they were repeated[ The designs were presented one at a time for one second each\ with each successive card covering the preceding one[

For the test phase\ one copy of each of the 05 target designs was randomly mixed with 05 distractor designs "i[e[\ frequency! 9#\ with the constraint that no two designs with the same fre! quency were presented consecutively[ The yes:no recognition and frequency judgement tests were administered immediately and concurrently after the study phase[ Subjects were given the following instructions] {{I|m now going to show you another set of cards consisting of the designs you|ve just seen\ plus some new ones that you haven|t seen[ For each design\ tell me if you saw it in the _rst set\ and if you did\ how many times you saw it[|| The subsequent analyses of performance on the estimation task were carried out only for those items that subjects correctly recognised from the study phase "i[e[\ true positives#[ We did not attempt to match recognition performance by varying the exposure times of the stimuli across the groups^ our method is modelled on that used by Smith and Milner ð22Ł and seeks to replicate and extend their _ndings across di}erent patient groups[

2[ Results A few subjects who scored close to ceiling "taken as greater than 89)# or ~oor "52) or less*{chance| per! formance  49)# on recognition memory were omitted from the analyses[ This left 02 diencephalic\ 09 temporal\ 02 frontal and 05 control subjects[ Mean recognition memory performance by these {mat! ched| groups was 65[4\ 65[7\ 79[7 and 70[7) for the diencephalic\ temporal\ frontal and control groups respectively^ these scores did not di}er signi_cantly\ F"2\37#  1[92\ P  9[01[ Recognition by these matched groups was also analysed in terms of d?[ Mean d? scores were 1[99\ 0[80\ 1[28 and 1[36\ which did not di}er sig! ni_cantly\ and the result was almost identical to the per! cent recognition scores\ F"2\37#  1[90\ P  9[01[ For all subsequent analyses the groups were matched in terms of d?[ Figure 0 shows frequency estimates across the subject groups\ plotted against {true| frequencies[ In order to

Fig[ 0[ Mean estimated frequency "and S[E[M[# of presentation of designs compared to true frequency*matched subjects[

N[ Stanhope et al[:Frequency judgements of abstract designs

determine whether the patients showed a dis! proportionate impairment in estimating frequency rela! tive to recognition performance\ a group x test "d? recognition:total estimates score# ANOVA was carried out[ There were highly signi_cant e}ects of group\ F"2\37#  8[85\ P ³ 9[9990\ and of test\ F"0\37#  166[86\ P ³ 9[9990[ The interaction was highly signi_cant\ F"2\37#  8[56\ P ³ 9[9990\ con_rming that although the groups were matched on recognition\ the patients were nonetheless disproportionately impaired in frequency judgements[ Having established that there was a disproportionate impairment in estimating frequency\ we carried out a two!way ANOVA "group x frequency# to examine group e}ects across the di}erent frequency levels[ This revealed a main e}ect of group\ F"2\37#  8[88\ P ³ 9[9990[ Post hoc NewmanÐKeuls tests showed that the three patient groups di}ered signi_cantly from healthy controls "P ³ 9[94#\ but that none of the patient groups di}ered from each other[ There was also a main e}ect of frequency\ F"2\033#  35[73\ P ³ 9[9990[ Consistent with previous studies ð0\ 23\ 25Ł\ all groups tended to under! estimate at higher frequencies[ An unexpected _nding was that controls overestimated at frequency!0\ although Begg ð0Ł reported that healthy subjects judged once!pre! sented words\ particularly abstract nouns\ as having been presented between three and four times[ There was a signi_cant group by frequency interaction\ F"8\033#  1[65\ P ³ 9[90[ As Fig[ 0 shows\ the control subjects| tendency to give higher estimates for high fre! quency items was less apparent in the patient groups[ Planned contrast analyses con_rmed this[ The combined patient groups were disproportionately impaired at mak! ing accurate estimates relative to the healthy controls "group by estimates interaction\ F"2\033#  5[31\ P ³ 9[990#[ Because we were interested in determining the e}ects of limbic!diencephalic lesions on performance\ we conducted a further analysis to see whether the dien! cephalic:temporal lobe patients di}ered from those with frontal lesions[ Comparison of the combined dien! cephalic and temporal lobe groups with the frontal pat! ients showed a marginal e}ect of group\ F"0\23#  2[31\ P ³ 9[96\ but no signi_cant group by estimates inter! action\ F  0[04[ Taken together\ these results indicate that the patterns of performance by the three patient groups were broadly similar "in that there was no inter! action#\ but that the diencephalic and temporal lobe pat! ients tended to be more severely impaired overall[ In the present study\ as in Smith and Milner|s ð22Ł and Strauss et al[|s ð24Ł studies\ progressively fewer items were shown at higher frequencies[ Consequently\ there was the opportunity for improved scores through practice on the low frequency items\ whereas high frequency items might be less reliable because the mean estimates were derived from a smaller number of trials[ We examined this possi! bility in three ways[ First\ at each frequency level we carried out an analysis of the results from the _rst trial only[ This gave a signi_cant main e}ect of group\

0280

F"2\37#  6[97\ P ³ 9[9994\ a signi_cant e}ect of frequency\ F"2\033#  22[83\ P ³ 9[9990\ and a sig! ni_cant group by frequency interaction\ F"8\033#  1[95\ P ³ 9[94[ On post hoc NewmanÐKeuls\ each of the pati! ent groups performed signi_cantly worse than controls "P ³ 9[94#\ but none of the patient groups di}ered sig! ni_cantly from one another[ Secondly\ we examined the _rst three trials across each of the 0\ 3 and 7 frequency levels[ This gave signi_cant main e}ects of group\ F"2\37#  7[69\ P ³ 9[9990\ and frequency\ F"1\85#  45[34\ P ³ 9[9990\ but the group by frequency interaction was not statistically signi_cant\ F"5\85#  0[58[ Again\ post hoc tests showed that each of the patient groups di}ered signi_cantly "P³9[94# from controls[ Thirdly\ we pooled the data from the 3\ 7 and 01 frequency levels "i[e[\ involving nine designs# and com! pared the mean estimates with the mean estimates from frequency!0 "i[e[\ seven designs# "compare Ref[ ð24Ł#[ There were signi_cant main e}ects of group\ F"2\37#  3[79\ P ³ 9[90\ and frequency\ F"0\37#  26[14\ P ³ 9[9990\ but the group by frequency interaction was not signi_cant\ F"2\37#  0[04[ In brief\ these analyses indicate that each of the patient groups gave signi_cantly lower estimates of frequency than did the healthy controls\ and this held good even after con! trolling for the number of trials at each frequency level[ However\ the group by frequency interaction was only obtained as a signi_cant _nding when frequency!01 was included as a separate data!point[ Analysis was carried out to compare performance across the six patient groups "Korsako}\ pituitary irradiation\ herpes encephalitis\ anoxic:epileptic tem! poral lobe lesions\ frontal tractotomy and focal frontal# and controls[ A group x test "d? recognition:total esti! mates score# ANOVA revealed a highly signi_cant e}ect of group\ F"5\34#  4[50\ P ³ 9[990\ and of test\ F"0\34#  066[97\ P³9[9990[ NewmanÐKeuls analysis of the group e}ect showed that the Korsako} and anox! ic:epileptic patients were worse than controls "P ³ 9[94#\ but that there were no other di}erences\ and\ in particu! lar\ the herpes encephalitis group did not di}er from the anoxic:epilepsy group[ There was a highly signi_cant interaction\ F"5\34#  4[12\ P ³ 9[990\ indicating that\ as with the three main lesion groups\ estimation per! formance by the patient subgroups was dis! proportionately impaired relative to recognition memory[ For frequency estimates taken in isolation\ a two!way ANOVA "group x frequency# showed a main e}ect of group\ F"5\34#  4[34\ P ³ 9[990\ and of frequency\ F"2\024#  22[73\ P ³ 9[9990\ and a signi_cant group x frequency interaction\ F"07\024#  1[92\ P ³ 9[90[ Plan! ned contrast analyses were carried out between the pati! ent subgroups[ All the patient subgroups\ except the pituitary patients\ di}ered from the control subjects in terms of the group main e}ect[ Comparison of the Kor! sako} group with the pituitary irradiation patients showed no e}ect of group "F  0[73#\ and the group

0281

N[ Stanhope et al[:Frequency judgements of abstract designs

Fig[ 1[ Mean estimated frequency "and S[E[M[# of presentation of designs compared to true frequency*Korsako} and pitu! itary patients[

x frequency interaction was not signi_cant "F  0[67#[ However\ Fig[ 1 shows a striking di}erence in the patterns of performance between these two groups\ in which the pituitary but not the Korsako} patients show a steady increase in frequency estimates with increasing true fre! quency[ In addition\ although the Korsako} group showed a group x frequency interaction when compared with controls\ the pituitary patients did not[ The failure to _nd signi_cant di}erences between the pituitary pat! ients and controls may be partly because of the small sample size\ but overall the results suggest that the dien! cephalic group|s impairment in making accurate esti! mates was very largely attributable to the Korsako} patients[ The analysis comparing the anoxic:epileptic pat! ients with the herpes encephalitis patients showed no main e}ect of group "F  9[20# and no interaction "F  9[63#[ The anoxic:epileptic patients showed a group x frequency interaction when compared with the controls "F  2[00\ P ³ 9[94#\ and the interaction was marginally signi_cant for the herpes encephalitis patients "F  1[41\ P ³ 9[95#[ The focal frontal patients did not di}er from the tractotomy patients in terms of the group e}ect "F9[57# or interaction "F  0[16#[ The relationship between d? recognition and frequency estimates was examined in a correlational analysis[ Mean estimates were calculated as a proportion of true fre! quency with the mean weighted according to the number of trials at each frequency level "i[e[\ weighted mean pro! portion#[ In the total patient group "N  25# and the temporal lobe patients separately\ there were signi_cant positive correlations "two!tailed# between d? recognition and estimates as a proportion of true frequency\ r  9[30\ P ³ 9[90\ and r  9[53\ P ³ 9[94\ respectively[ There were no signi_cant correlations in the diencephalic\ r  9[96\ frontal\ r  9[05\ or control groups\ r  9[00[ Because of the association between d? recognition and frequency estimates in the total patient group and tem! poral lobe group\ we partialled out d? recognition to

control for this factor in subsequent correlational analy! ses[ Two!tailed pairwise correlations were conducted between frequency estimates "weighted mean proportion# and current and premorbid IQ\ age\ GMQ and DMQ\ and frontal:executive tests\ using the data from the total patient group\ and individual correlations on the data from the three patient groups and the control group[ Because of the large number of calculations\ a con! servative criterion of P ³ 9[91 was selected[ There were no signi_cant correlations between estimates and these variables in the expected direction in either the total pati! ent group or the individual subject groups[ If it is assumed that frequency is an aspect of context memory\ one might expect performance on the present task to correlate with performance on other tests of con! text memory[ In another study ð04Ł we examined the ability in the same patients to make temporal and spatial discriminations[ Subjects were shown two sets of 04 line drawings of objects\ each presented at the top or the bottom of a ~ash card^ the second set was presented 34 min after the _rst set\ and retention was tested 09 min after presentation of the second set[ Subjects were asked to try to remember the objects\ whether each was at the top or bottom of the card "spatial discrimination# and which of the two sets "_rst or second# each object had appeared in "temporal discrimination#[ The data from the matched groups in the present study "N  25# were analysed to determine whether there was a relationship between frequency estimates "weighted mean proportion\ as above# and their ability to make temporal and spatial discriminations "proportion of items correctly recognised in which the correct context judgement was also made\ using Hunkin\ Parkin and Longmore|s ð7Ł z correction for di}erential hit rates#[ Table 2 shows the results of the pairwise correlations "two!tailed#[ In the total patient group and the temporal lobe group analysed individually\ spatial discrimination correlated with frequency esti! mates^ i[e[\ better spatial discrimination was associated with a higher proportion of correct frequency estimates[ The diencephalic group showed a trend in the same direc! tion[ There were no correlations between frequency esti! mates and temporal discrimination in either the total patient group or when the groups were analysed separ!

Table 2[ Correlations between frequency estimates "weighted mean proportion# and performance on temporal and spatial discrimination

All patients "N25# Diencephalic Temporal Frontal Control

Temporal discrimination

Spatial discrimination

9[06 9[09 9[38 −9[98 −9[98

9[33 9[28 9[52 9[19 −9[29

 P ³ 9[94^  P ³ 9[91[

N[ Stanhope et al[:Frequency judgements of abstract designs

ately\ although the temporal lobe group showed a trend towards a positive association between the two[ Further analysis using Fisher|s Zr transformation was conducted to see whether the correlations between esti! mation performance and spatial:temporal discrimination di}ered between groups ð3Ł[ For spatial discrimination\ the correlation coe.cient for the total patient group di}ered signi_cantly from that of the controls\ z  1[28\ P ³ 9[94[ Likewise\ the correlation for temporal lobe pat! ients also di}ered from controls\ z  1[10\ P ³ 9[94[ There were no signi_cant group di}erences in the cor! relation coe.cients between temporal discrimination and frequency estimates[ Within the subject groups\ none of the frequency:spatial and frequency:temporal cor! relations di}ered signi_cantly[

Discussion Previous investigations of memory for context and fre! quency estimations have yielded inconsistent results[ On the one hand\ several studies have presented data clai! ming that de_cits in memory for a range of contextual attributes including frequency estimates are due to frontal lobe dysfunction\ e[g[\ ð17\ 25Ł[ On the other hand\ a number of studies have suggested that such de_cits may re~ect a more fundamental {core| memory impairment rather than frontal problems\ e[g[\ ð01\ 13\ 14\ 20Ł[ This study has demonstrated that despite comparable levels of item recognition\ diencephalic and temporal lobe patients and those with frontal lobe damage have more di.culty than healthy controls in estimating how often previously presented designs appeared[ The diencephalic and temporal lobe patients were similarly and severely impaired\ and in the former there was a non!signi_cant di}erence in the pattern of performance between the pitu! itary and Korsako} patients[ Although these two sub! groups did not di}er signi_cantly in recognition\ the Korsako} patients| degree of underestimation was strik! ing^ they showed little increase in the estimates they made\ such that those made to frequency!0 and frequency!01 items di}ered by an absolute value of only 9[82[ The impairment shown by the diencephalic group compared with controls therefore arose mainly because of the Kor! sako} patients| signi_cant de_cits and indicates that di.culty judging the frequency of occurrence was not uniform within the diencephalic group[ The di}erence between the Korsako} and pituitary patients may result from two factors[ First\ there is a greater likelihood of frontal involvement in the Korsako} patients\ and second\ in terms of absolute level of memory impairment "as opposed to IQ!MQ di}erence#\ the Korsako} group were more severely amnesic[ There was a signi_cant correlation between frequency estimates and item recognition memory in the total pati! ent group and in the temporal lobe group when analysed separately[ Estimates did not correlate signi_cantly with any measures of frontal lobe function in either the total

0282

group or individual patient groups[ When considered together\ the _ndings suggest that de_cits in frequency estimation ability may arise for di}erent reasons in the patient groups[ In the temporal lobe patients the de_cit may re~ect a core memory impairment rather than a speci_c impair! ment of this aspect of context memory[ This suggestion is consistent with other studies reporting context memory de_cits which appear to be associated more with target memory performance than with frontal lobe function ð01\ 13\ 14\ 20Ł[ The estimation de_cits shown by the temporal lobe patients can be explained in terms of the multiple! trace model ð5\ 6Ł[ According to this\ each presentation of an item produces a new memory trace[ Frequency judgements are then made by counting the number of traces that can be retrieved[ In a patient who has a gen! erally poor memory\ some of those traces may be too weak to be retrieved and so result in an underestimate "also see ð8Ł#[ Alternatively\ subsequent presentations of a design may strengthen earlier representations of that design\ so that stronger\ but fewer traces are formed[ In either case\ the item can be recognised at test but there will be a de_cit in judging how often it has been seen[ The results from our temporal lobe group initially appear inconsistent with those of Sagar et al[ ð15Ł[ Those authors found that patient H[M[ showed normal frequency dis! crimination even when his recognition was extremely poor[ However\ their study and ours are not directly comparable in various ways[ In particular\ H[M[ was required to judge between the relative frequency of two previously!presented words and not asked to estimate their absolute frequency[ While the de_cits shown by the frontal lobe patients may also be to some extent due to an anterograde mem! ory de_cit relative to controls\ there may be an alternative "or additional# explanation for their defective perform! ance[ Luria ð05Ł suggested that even though memory func! tion in frontal lobe patients may be intact\ other processes crucial to memory may be impaired[ Recent work sup! ports this\ and shows that frontal lesions frequently lead to di.culty in planning\ monitoring and organising infor! mation in memory\ e[g[\ ð25Ł[ If it is assumed that each presentation of a given design produces its own memory trace\ frontal lobe patients| defective performance may re~ect their di.culties in making an organised and e}ec! tive search through memory to detect multiple rep! resentations of the same design\ even though performance did not correlate signi_cantly with scores on executive tests[ Their di.culty on the present task may re~ect a general problem in making precise numerical estimates[ However\ their judgements were reasonably accurate at the lower frequencies "0 and 3# compared with their marked underestimation at higher frequencies "7 and 01#[ This suggests that\ if several instances of an item are to be identi_ed and their number accurately estimated\ it is particularly important to adopt an orderly search strat! egy^ the ability to do so may require unimpaired frontal lobe function[

0283

N[ Stanhope et al[:Frequency judgements of abstract designs

The explanation for the diencephalic patients| impaired performance is more speculative[ Within the diencephalic group as a whole\ there were trends for those patients with the most severe memory impairment and the poorest performance on card!sorting to show the lowest fre! quency estimations in terms of total scores "correlation with DMQ] r  9[43\ P ³ 9[95\ and with card!sorting perseverations] r  −9[33\ NS within this small group#[ The diencephalic group comprised patients with Kor! sako}|s syndrome and the pituitary patients\ and the performance of these two subgroups di}ered qualitatively from each other "see Fig[ 1#[ Compared with the pituitary group\ the Korsako} patients had a variable degree of cortical atrophy\ as well as frontal hypometabolism in some cases\ as shown by our PET data "see above#[ Consequently\ as in the case of the frontal lobe patients\ the Korsako} patients| frontal pathology may have led to di.culty in implementing an e}ective search for those memory traces that were formed[ Secondly\ the Kor! sako} group|s memory impairments were more severe than those of the pituitary group in terms of the absolute level of their memory quotients[ The GMQ and DMQ of the Korsako} group and the temporal lobe patients were very similar\ suggesting an equivalent degree of memory impairment in these two groups[ Consequently\ a mul! tiple!trace explanation "i[e[\ weak traces\ or stronger but fewer traces# would also be consistent with the Korsako} patients| de_cits[ In brief\ therefore\ the Korsako} patients| poor estimation ability may be due to the combined e}ects of frontal pathology and a generally poor memory[ There were signi_cant correlations between spatial dis! crimination and estimation performance in the total pati! ent group and temporal lobe patients and a trend in the diencephalic group\ as well as a trend in the temporal lobe group between temporal discrimination and estimation performance[ These _ndings provide some support for the view that frequency estimation is a form of context memory[ In other words\ the ability to perform the dis! crimination and estimation tasks relies to some extent on similar processes\ and successful performance may depend\ at least in part\ on the ability to retrieve the speci_c contextual attributes of a stimulus that allow it to be distinguished from other stimuli[ The similarities and di}erences between our _ndings and those reported by Smith and Milner ð22Ł should be considered[ Smith and Milner found that all their subjects tended to underestimate at higher frequency levels[ While this was also the case in the present study\ the interaction we obtained re~ects the fact that the group di}erences were more marked at higher frequencies than in their study[ As in our study\ Smith and Milner|s frontal lobe group showed impairment at the higher frequency levels[ Smith and Milner also suggested that the asymptote observed in their frontal patients above frequency!6 might emerge in healthy controls at higher frequencies[ The data from our controls suggest that if this is the case\ the control asymptote would not occur until frequencies

beyond 01[ Consistent with this is Begg|s ð0Ł _nding that healthy control subjects had not reached asymptote even with words presented 06 times[ In addition\ frequency judgements in Smith and Milner|s temporal lobe patients were better than those made by our temporal lobe group\ which\ like our Korsako} patients\ reached asymptote by frequency!7[ In our study\ the temporal lobe group performed marginally worse than the frontal lobe group\ but the di}erence was not statistically signi_cant[ In their study\ the frontal groups performed marginally worse than the temporal lobe groups\ but\ again\ the di}erences appear not to have been statistically signi_cant[ Methodological di}erences between the two studies may account for those results which were discrepant[ Smith and Milner presented 19 di}erent designs in the study phase\ whereas we presented 05 designs in the study phase and included 05 distractors at test\ against only 3 distractors used by Smith and Milner[ It is possible that the di}erent proportions between targets and distractors at test in the two studies a}ected the sensitivity of the task and thereby subjects| performance*their temporal lobe patients might have shown signi_cant de_cits in estimation ability had there been more distractor items in their study[ Control subjects| estimates in our study were also lower than in Smith and Milner|s\ and again this may indicate the di}erential sensitivity of the two studies[ Lastly\ di}erences between the type of lesion in Smith and Milner|s patients and in ours may account for those results which were inconsistent[ Their patients had all undergone unilateral temporal or frontal lobe excision to relieve focal epileptic seizures\ whereas our temporal lobe patients were a mixed group who were memory! impaired as a result of either herpes encephalitis\ anoxia or epilepsy[ In brief\ our results may di}er from Smith and Milner|s for two reasons[ First\ our temporal lobe patients were probably more memory impaired[ Second\ Smith and Milner|s task was easier\ and therefore argu! ably less sensitive[ In summary\ the present study demonstrates that de_! cits in making frequency judgements can arise directly from temporal lobe lesions\ circumscribed frontal lobe damage\ or diencephalic lesions[ In the temporal lobe group\ this is related to the severity of their target memory impairment[ In the frontal group\ the pathology may impair the ability to make organised searches in memory for multiple instances of an item[ The evidence regarding the underlying di.culty in diencephalic patients is less conclusive\ but a comparison between the Korsako} and pituitary patients suggests that the Korsako} group|s de_cit may be due at least partly to both general memory impairments and to frontal dysfunction[

Acknowled`ements*This research was funded on a Wellcome Trust grant to Drs M[ Kopelman\ J[ Wade and B[ Kendall[ The authors thank Dr L[ Reed and Mr D[ Lasserson for help in scan analysis[

N[ Stanhope et al[:Frequency judgements of abstract designs

References 0[ Begg I[ Estimation of word frequency in continuous and discrete tasks[ Journal of Experimental Psy! chology 0863^091]0935Ð41[ 1[ Benton AL[ Di}erential behavioral e}ects of frontal lobe disease[ Neuropsychologia 0857^5]42Ð59[ 2[ Butters N\ Salmon DP\ Munro Cullum D\ Cairns P\ Troster AI\ Jacobs D[ Di}erentiation of amnesic and demented patients with the Wechsler Memory Scale*Revised[ The Clinical Neuropsychologist 0877^1]022Ð37[ 3[ Ferguson GA[ Statistical Analysis in Psychology and Education[ 1nd ed[ London] McGraw!Hill\ 0855[ 4[ Guinan EM\ Kopelman MD\ Stanhope N\ Lewis P\ Lowy C[ The cognitive e}ects of pituitary adenomas and their treatments] two case studies and an inves! tigation of 89 patients[ Journal of Neurology\ Neuro! surgery and Psychiatry in press[ 5[ Hintzman DL\ Block RA[ Repetition and memory] evidence for a multiple!trace hypothesis[ Journal of Experimental Psychology 0860^77]186Ð295[ 6[ Hintzman DL\ Grandy CA\ Gold E[ Memory for frequency] a comparison of two multiple trace theor! ies[ Journal of Experimental Psychology] Human Learning and Memory 0870^6]120Ð39 7[ Hunkin NM\ Parkin AJ\ Longmore BE[ Aetiological variation in the amnesic syndrome] comparisons using the list discrimination task[ Neuropsychologia 0883^21]708Ð14[ 8[ Huppert FA\ Piercy M[ The role of trace strength in recency and frequency judgements by amnesics and control subjects[ Quarterly Journal of Experimental Psychology 0867^29]235Ð43[ 09[ Janowsky JS\ Shimamura AP\ Squire LR[ Source amnesia impairment in patients with frontal lobe lesions[ Neuropsychologia 0878^16]0932Ð45[ 00[ Kartsounis LD\ Poynton A\ Bridges PK\ Bartlett FR[ Neuropsychological correlates of stereotactic subcaudate surgery[ Brain 0880^003]1546Ð62[ 01[ Kopelman\ M[D[ Remote and autobiographical memory\ temporal context memory\ and frontal atrophy in Korsako} and Alzheimer patients[ Neu! ropsychologia 0878^16]326Ð59[ 02[ Kopelman MD[ The Korsako} syndrome[ British Journal of Psychiatry 0884^055]043Ð62[ 03[ Kopelman MD[ What|s wrong and what|s right in neuropsychiatry and neuropsychiatry[ Cognitive Neuropsychiatry 0885^0]164Ð77[ 04[ Kopelman MD\ Stanhope N\ Kingsley D[ Temporal and spatial context memory in patients with focal diencephalic\ temporal lobe\ and frontal lobe lesions[ Neuropsychologia 0886^24]0422Ð34[ 05[ Luria AR[ The working brain] An introduction to neuropsychology[ New York] Basic Books\ 0862[ 06[ Mayes AR\ Baddeley AD\ Cockburn J\ Meudell PR\ Pickering A\ Wilson B[ Why are amnesic judgements of recency and frequency made in a qualitatively di}erent way from those of normal people< Cortex 0878^14]368Ð77[ 07[ Meudell PR\ Mayes AR\ Ostergaard A\ Pickering A[ Recency and frequency judgements in alcoholic

08[ 19[ 10[ 11[

12[

13[ 14[ 15[ 16[ 17[

18[ 29[

20[ 21[ 22[

23[

24[

25[

0284

amnesics and normal people with poor memory[ Cor! tex 0874^10]376Ð400[ Milner B[ Some cognitive e}ects of frontal!lobe lesions in man[ Philosophical Transactions of the Royal Society of London\ B 0871^187]100Ð15[ Nelson HE[ A modi_ed card!sorting test sensitive to frontal lobe de_cits[ Cortex 0865^01]202Ð13[ Nelson HE\ Willison JR[ The National Adult Read! ing Test[ Second Edition\ Windsor] NFER!Nelson\ 0880[ Parkin AJ\ Hunkin NM[ Impaired temporal context memory on anterograde but not retrograde tests in the absence of frontal pathology[ Cortex 0882^18]156Ð79[ Parkin AJ\ Leng NRC\ Stanhope N\ Smith A[P[ Memory impairment following ruptured aneurysm of the anterior communicating artery[ Brain and Cognition 0877^6]120Ð32[ Parkin AJ\ Leng NRC\ Hunkin N[ Di}erential sen! sitivity to contextual information in diencephalic and temporal lobe amnesia[ Cortex 0889^15]262Ð79[ Pickering A\ Mayes AR\ Fairbairn AF[ Amnesia and memory for modality information[ Neu! ropsychologia 0878^16]0138Ð48[ Sagar\ HJ\ Gabrieli JDE\ Sullivan EV\ Corkin S[ Recency and frequency discrimination in the amnesic patient H[M[ Brain 0889^002]470Ð591[ Scoville WB\ Milner B[ Loss of recent memory after bilateral hippocampal lesions[ Journal of Neurology\ Neurosurgery and Psychiatry 0846^19]00Ð10[ Schacter DL\ Harbluk JL\ McLachlan DR[ Retrieval without recollection] An experimental analysis of source amnesia[ Journal of Verbal Learning and Ver! bal Behavior 0873^12]482Ð500[ Shallice T\ Evans ME[ The involvement of the frontal lobes in cognitive estimates[ Cortex 0867^03]183Ð292[ Shimamura AP\ Janowsky JS\ Squire LR[ Memory for the temporal order of events in patients with frontal lobe lesions and amnesic patients[ Neu! ropsychologia 0889^17]792Ð02[ Shoqeirat MA\ Mayes AR[ Disproportionate inci! dental spatial!memory and recall de_cits in amnesia[ Neuropsychologia 0880^18]638Ð58[ Smith ML\ Milner B[ E}ects of focal brain lesions on sensitivity to frequency of occurrence[ Society for Neuroscience Abstracts 0872^8]29[ Smith ML\ Milner B[ Estimation of frequency of occurrence of abstract designs after frontal or tem! poral lobectomy[ Neuropsychologia 0877^15]186Ð 295[ Squire LR[ Comparisons between forms of amnesia] some de_cits are unique to Korsako}|s syndrome[ Journal of Experimental Psychology] Learning\ Memory and Cognition 0871^7]459Ð60[ Strauss ME\ Weingartner H\ Thompson K[ Remem! bering words and how often they occurred in mem! ory!impaired patients[ Memory and Cognition 0874^02]496Ð09[ Stuss DT\ Alexander MP\ Palumbo CL\ Buckle L\ Sayer L\ Pogue J[ Organizational strategies of pat! ients with unilateral or bilateral frontal lobe injury in word list learning tasks[ Neuropsychology 0883^7]244Ð62[

0285

N[ Stanhope et al[:Frequency judgements of abstract designs

26[ Victor M\ Adams RD\ Collins GH[ The WernickeÐ Korsako} syndrome[ A clinical and pathological study of 134 patients\ 71 with post!mortem exam! inations[ Oxford] Blackwell\ 0860[ 27[ Von Cramon DY\ Markowitsch HJ[ The problem of localising memory in focal cerebrovascular lesions[

In] Neuropsychology of Memory "1nd Edn#\ Squire LR\ Butters N\ editors[ New York\ Guilford Press\ 0881\ pp[ 84Ð094[ 28[ Wechsler\ D[ Wechsler Adult Intelligence Scale* Revised[ Psychological Corporation\ London and New York\ 0870[