Accelerated long-term forgetting in children with idiopathic generalized epilepsy


  • Michael B. Gascoigne,

    1. School of Psychology, and ARC Centre of Excellence in Cognition and its Disorders, The University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Belinda Barton,

    1. CHERI, The Children’s Hospital at Westmead and Discipline of Paediatrics and Child Health, Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Richard Webster,

    1. T.Y. Nelson Department of Neurology and Neurosurgery, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Deepak Gill,

    1. T.Y. Nelson Department of Neurology and Neurosurgery, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Jayne Antony,

    1. T.Y. Nelson Department of Neurology and Neurosurgery, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Suncica Sunny Lah

    1. School of Psychology, and ARC Centre of Excellence in Cognition and its Disorders, The University of Sydney, Sydney, Australia
    Search for more papers by this author

Address correspondence to Suncica S. Lah, School of Psychology, University of Sydney, Sydney, NSW 2006, Australia. E-mail:


Purpose:  The rapid forgetting of information over long (but not short) delays (accelerated long-term forgetting [ALF]) has been associated with temporal lobe epilepsy but not idiopathic generalized epilepsy (IGE). Long-term memory formation (consolidation) is thought to demand an interaction between medial temporal and neocortical networks, which could be disrupted by epilepsy/seizures themselves. The present study investigates whether ALF is present in children with IGE and whether it relates to epilepsy severity.

Methods:  Sixty-one children (20 with IGE and 41 healthy controls [HC]) of comparable age, sex, and parental socioeconomic status completed neuropsychological tests, including a measure of verbal learning and recall after, short (30-min) and long (7-day) delays, and recognition. Epilepsy severity was rated by treating neurologists.

Key Findings:  A two-way repeated measures analysis of covariance (ANCOVA) found a significant Group x Delay interaction; the children with IGE recalled (and recognized) significantly fewer words after a long, but not short (2- and 30-min) delay relative to the HC children. Moreover, greater epilepsy severity was associated with poorer recognition.

Significance:  This study demonstrates, to our knowledge for the first time, that children with IGE present with ALF, which is related to epilepsy severity. These findings support the notion that epilepsy/seizures themselves may disrupt long-term memory consolidation, which interferes with day-to-day functioning of children with IGE.

Memory difficulties represent the most common clinical complaints in patients with epilepsy (Corcoran & Thompson, 1992; Vermeulen et al., 1993; Blake et al., 2000). Yet both adult (Elixhauser et al., 1999; Piazzini et al., 2001) and child (Kadis et al., 2004) studies suggest that these memory complaints are often unrelated to scores obtained on standardized memory tests, and are not confined to patients with temporal lobe epilepsy. This paradoxical lack of correspondence between memory complaints and neuropsychological test scores may (in part) be due to overly short delays used on the tests. Of interest, some patients with TLE have difficulties recalling information after long delays despite intact recall on short delays used on neuropsychological tests; they exhibit accelerated long-term forgetting (ALF) which has been attributed to hippocampal pathology/seizure focus (Martin et al., 1991; Kapur et al., 1997; O’Connor et al., 1997; Blake et al., 2000; Mameniskiene et al., 2006; Jansari et al., 2010; Muhlert et al., 2011; Wilkinson et al., 2012). However, the role of the hippocampus in ALF is not proven as several studies that involved patients with TLE failed to find evidence of ALF (Giovagnoli et al., 1995; Bell et al., 2005; Bell, 2006; Howard et al., 2010), and in patients with transient epileptic amnesia (TEA) hippocampal atrophy did not correlate with ALF (Butler et al., 2009).

It has been proposed that long-term memory formation requires a consolidation process that may take hours, days, or months to complete (Squire et al., 1984), and demands an interaction between the hippocampus and neocortex (Squire & Alvarez, 1995; Mayes et al., 2003). Therefore, not only hippocampal pathology, but also disruption (such as from seizure activity) of the transfer of information between the hippocampus and neocortex during the memory consolidation could cause ALF (Mayes et al., 2003). Therefore, ALF could be found in patients with generalized seizures who have no detectable cortical pathology: idiopathic generalized epilepsy (IGE; Davidson et al., 2007).

To date, two studies have examined ALF in IGE populations: one involved adults (Muhlert et al., 2011) and another children (Davidson et al., 2007). The former found no evidence of memory deficits on short (30-min) and long delays (3 weeks) in patients with IGE. Findings of adult studies, however, should not be automatically generalized to children, as children’s cognitive skills and functional brain organization are still developing. Indeed, in children with IGE Davidson et al. (2007) found evidence of ALF, but attributed this finding to poor initial learning efficiency. Although seizures were proposed to play a critical role in long-term consolidation, 24% of the sample of Davidson et al. had been seizure free for 2 years, and the adult study did not state whether seizure-free participants were included. Inclusion of seizure-free participants could have reduced the likelihood of finding ALF in these studies. Nevertheless, adult TLE studies have yielded inconsistent findings on the role of seizures in ALF: two studies found significant associations between seizures and rate of forgetting over long delays (Mameniskiene et al., 2006; Wilkinson et al., 2012), and one study did not (Blake et al., 2000). It is possible that instead of a single factor, a combination of epilepsy-related factors play a significant role in ALF, which could account for inconsistent findings. For example, greater seizure severity (Mameniskiene et al., 2006), longer duration of epilepsy disorder (Blum, 2001; Nolan et al., 2004; Wang et al., 2011), and the use of poly-pharmacy (Alessio et al., 2004) were all found to be associated with poor learning and recall after short delays. Therefore, a measure that considers overall epilepsy severity, rather than epilepsy factors separately, could be related to the rate of long-term forgetting.

This study aimed to examine long-term verbal memory formation in children with IGE and determine whether it relates to epilepsy severity and day-to-day memory difficulties. To overcome the shortcomings of previous studies we included only children who had experienced at least one overt seizure in the last 2 years and examined the relationship between epilepsy and long-term memory formation using a validated epilepsy severity scale. We hypothesized that children with IGE will have poorer recall and recognition of verbal information after long, but not short delays, and that poorer long-term memory will be associated with more severe epilepsy and everyday memory complaints.



There were 20 children diagnosed with IGE and 41 healthy control (HC, free of epilepsy history) participants aged 6–16 years, fluent in English, with Full Scale Intelligence Quotient (FSIQ) >79, and free of preexisting diagnoses of major: sensory, neurodevelopmental [but not Attention Deficit Hyperactivity Disorder (ADHD) or learning difficulties] and other neurological disorders. Children with IGE were recruited through the neurology department, The Children’s Hospital at Westmead (CHW). Medical history, electroencephalography, and imaging data were reviewed by treating neurologists according to the International League Against Epilepsy (ILAE) criteria for IGE (Commission on Classification and Terminology of the International League Against Epilepsy, 1989). Only children who had experienced at least one seizure in the 2-year period before the assessment and whose clinical magnetic resonance imaging (MRI) scans (when available, n = 14) were free of structural abnormalities were considered for the study; 27 (87%) of 31 consented. Seven were excluded because subsequent findings disqualified IGE diagnosis (n = 6), or long-term verbal memory materials were noted and rehearsed, invalidating the test (n = 1). Control participants were recruited via word-of-mouth (snowball recruitment), through the peer networks of both IGE and other control participants.


Socioeconomic status (SES)

Daniel’s Scale of Occupational Prestige (Daniel, 1983) rates parental occupation on a seven-point scale, with higher scores reflecting lower SES. Mean parental occupation scores were used, except for single-parent families.

Neuropsychological measures

Clinical neuropsychological instruments are summarized in Table 1.

Table 1.   Clinical neuropsychological instruments
DomainTest: SubtestScore used
  1. WASI, Wechsler Abbreviated Scale of Intelligence (Wechsler, 1999); EMQ, Everyday Memory Questionnaire (Kadis et al., 2004); WISC-IV, Wechsler Intelligence Scale for Children, Fourth Edition (Wechsler, 2003); WMTB-C, Working Memory Battery for Children (Pickering & Gathercole, 2001); WRAML2, Wide Range Assessment of Memory and Learning, Second Edition (Sheslow & Adams, 2003); CMS, Children’s Memory Scale (CMS) (Cohen, 1997).

IntelligenceWASI: Vocabulary and MatricesFull Scale Intelligence Quotient (FSIQ)
M = 100, SD = 15
Everyday memoryEMQRaw score
Working memoryWISC-IV: Digits SpanAge scaled scores
M = 10, SD = 3
WMTB-C: Block RecallAge scaled scores
M = 10, SD = 3
Short-term memoryWRAML2: Story MemoryAge scaled scores
M = 10, SD = 3
CMS: Dot LocationAge scaled scores
M = 10, SD = 3

Experimental verbal memory measure

Word lists were derived from the California Verbal Learning Test – Children’s Version (CVLT-C, Delis et al., 1994), which has been shown to have acceptable reliability (coefficient alpha ranging from 0.72 to 0.85; Delis et al., 1994) and construct validity (Griffiths et al., 2006) in children with epilepsy. A 9- and 12-word list was devised for children <9 and ≥9 years of age, respectively. First, this list (A) was read until all words were recalled on two consecutive trials or 12 learning trials (two IGE participants) were completed. Second, list B (interference) was read and recalled once. Third, a recall of list A was requested following two short (2- and 30-min) and one long (one week) delay, without forewarning. Fourth, on completion of the long delay recall, a recognition list was presented. Long delay recall and recognition were conducted via telephone. The following scores were used: (1) number of learning trials (learning efficiency), (2) percentage of words recalled (max 100%) relative to last learning trial (recall), and (3) d prime (d′, recognition), which takes into account the proportion of correctly and incorrectly recognised words (Macmillan & Creelman, 1991).

Epilepsy severity

The Global Assessment of Severity of Epilepsy (GASE, Speechley et al., 2008) rates overall epilepsy severity on a scale from 1 (Not at all severe) to 7 (Extremely severe). It considers the frequency and intensity of seizures, injuries during seizures, severity of postictal period, number and side effects of antiepileptic drugs, and interference of epilepsy or drugs with daily activities. This scale has good construct validity as well as high interrater (weighted κ = 0.85) and test–retest (weighted κ = 0.90) reliability.


The study was approved by the CHW and University of Sydney Ethics Research Committees. Parents gave informed consent and children were tested individually over one 90-min session and contacted 7 days later via phone. Memory questionnaires were completed by parents. Developmental history, SES, and information about relevant epilepsy variables were gathered via parental interview. Information about epilepsy was verified by treating pediatric neurologists and review of medical records. Neurologists, who were blinded to children’s scores on neuropsychological measures, completed the GASE scale.

Statistical analysis

All tests were two-tailed, with α < 0.05. Groups were compared using independent t-tests and chi-square tests on background variables, and ANCOVA (with centered FSIQ as a covariate) on other measures. Spearman’s rho (rs) was used to assess strengths of correlations between epilepsy severity and everyday memory with long-term and short-term memory measures.

Scores obtained on the experimental measure of verbal memory were analyzed using a two-way (Group × Delay) repeated measures ANCOVA, followed by Sidak post hoc tests. A priori hypotheses concerning ALF were assessed with the following planned contrasts; the 30-min and 7-day, and the 2- and 30-min delays.


IGE participants were diagnosed at a mean age of 6.09 (standard deviation, SD 3.25) years and had experienced epilepsy for an average of 4.54 (SD 2.18) years (Table S1). All IGE children were taking antiepileptic drugs (AEDs): 18 monotherapy and 2 polytherapy. The mean epilepsy severity score (M 2.17, SD 1.69) corresponded to a rating between “A little severe” and “Somewhat severe.” The IGE and control groups did not differ in age, SES, sex distribution, and frequency of ADHD or learning difficulties diagnosis (see Table 2). Mean FSIQ of the IGE group was significantly lower than that of the control group.

Table 2.   Demographic and intelligence variables
 IGE (n = 20)
Mean (SD)
Control (n = 41)
Mean (SD)
Test of significanceP-value
  1. IGE, idiopathic generalized epilepsy; SES, socioeconomic status; FSIQ, Full-Scale Intelligence Quotient; ADHD, attention deficit hyperactivity disorder; LD, learning difficulty.

Age10.76 (2.47)11.23 (2.63)t59 = −0.680.50
Sex (F/M)10/1021/20χ2 = 0.010.93
SES4.08 (1.31)3.70 (1.08)t59 = 1.130.27
FSIQ102.00 (10.60)111.32 (11.22)t59 = −3.10<0.01
Comorbid ADHD/LD1/00/0χ2 = 2.100.15

On the Everyday Memory Questionnaire (EMQ) ANCOVA revealed significantly lower parental ratings in the IGE compared to the control group (Table 3). No significant differences were found between the groups on tests of working or visual memory. However, on the verbal memory test (WRAML2: Story Memory) the IGE children recalled a significantly fewer details immediately after the presentation and on 30-min delay relative to the control group.

Table 3.   Scores obtained on clinical neuropsychological instruments
 IGE (n = 20)Control (n = 41)Test of significancep-value
Unadjusted mean (SD)Adjusted means (SE)Unadjusted means (SD)Adjusted means (SE)
  1. WMTB-C, Working Memory Test Battery for Children, (Pickering & Gathercole, 2001); WISC-IV, Wechsler Intelligence Scale for Children, Fourth Edition, (Wechsler, 2003); WRAML2, Wide Range Assessment of Memory and Learning, Second Edition, (Sheslow & Adams, 2003); CMS, Children’s Memory Scale, (Cohen, 1997).

  2. aRaw Scores, IGE, idiopathic generalized epilepsy; EMQ, Everyday Memory Questionnaire (Kadis et al., 2004).

EMQa72.84 (12.31)73.35 (3.16)87.29 (13.21)87.06 (2.09)F1,59 = 12.30<0.01
 Block recall93.95 (15.18)97.05 (4.64)93.63 (22.32)92.12 (3.16)F1,60 = 0.720.40
WISC-IV digit span      
 Digits forward8.85 (2.52)9.33 (0.59)10.95 (2.66)10.72 (0.40)F1,60 = 3.570.64
 Digits backward8.70 (2.31)9.01 (0.64)9.95 (2.92)9.80 (0.44)F1,60 = 0.990.32
WRAML-2 story memory      
 Immediate recall9.85 (2.89)10.34 (0.51)12.10 (1.97)11.86 (0.35)F1,60 = 5.740.02
 Delayed recall (30 min)9.30 (2.94)9.75 (0.52)11.83 (1.95)11.61 (0.35)F1,60 = 8.320.01
CMS Dot locations      
 Immediate recall9.85 (3.41)10.33 (0.75)10.17 (3.20)9.94 (0.51)F1,60 = 0.180.68
 Delayed recall (30 min)10.70 (3.18)10.98 (0.71)10.46 (2.93)10.33 (0.48)F1,60 = 0.560.46

Experimental measure of verbal memory

ANCOVA yielded no significant difference between the IGE (M 6.57, SD 2.59) and control (M 5.85, SD 2.15) groups in the number of trials needed to reach the learning criterion (F1,60 = 0.39, p = 0.54).

Recall and recognition data are presented in Fig. 1A,B, respectively. A two-way repeated-measures ANCOVA found a significant interaction between Group and Delay (F2,116 = 8.39, p < 0.01). Planned contrasts revealed that, compared to the control group, the IGE group recalled significantly fewer words from 30 min to 7 days (F1,58 = 9.89, p = 0.003), but not from 2 to 30 min (F1,58 = 0.74, p = 0.39). In addition, a main effect of Delay (F2,116 = 81.08, p < 0.001), but not Group (F1,58 = 1.36, p = 0.25) was evident. No significant difference in recall was found between the 2- and 30-min delays. However, participants recalled fewer words from the 30-min to 7-day delay (p < 0.001). Finally, the ANCOVA indicated that IGE participants had significantly poorer verbal recognition (d′) compared to control participants (F1,60 = 17.82, p < 0.001).

Figure 1.

Mean (standard error bars) recall (A) and recognition (B) scores obtained by the IGE and control groups on the experimental measure of verbal memory covaried for FSIQ.

Is ALF related to incomplete learning and poor verbal memory (story memory)?

To address this question, a repeated-measures ANCOVA that only included participants who learned all words (IGE, n = 18; HC, n = 41) was rerun, with Story Memory (immediate and delayed) and FSIQ as covariates. This ANCOVA again yielded a significant interaction between Group and Delay (F2,108 = 7.96, p = 0.001) and a significant main effect of Delay (F2,108 = 74.84, p < 0.001), but not Group (F1,54 = 2.60, p = 0.11). Planned contrasts showed that the IGE group recalled significantly fewer words from 30-min to 7-day (F1,54 = 10.81, p = 0.002), but not from 2- to 30-min (F1,54 = 0.21, p = 0.65) delays relative to the control group. Across groups, no differences were found between the number of words recalled on 2- and 30-min delays; however, less information was recalled from the 30-min to 7-day delay (p < 0.001). Finally, IGE participants had significantly worse recognition (d′) compared to control participants (F1,56 = 23.13, p < 0.001).

Relations between experimental memory measure, epilepsy severity, and everyday memory

A significant negative correlation was found between epilepsy severity ratings on the GASE scale and long-term verbal recognition accuracy (d′: rs = −0.49, p < 0.05); greater epilepsy severity was associated with worse long-term verbal recognition. However, epilepsy severity did not correlate with recall of information at any delay (2-min: rs = −0.10; 30-min: rs = −0.18; 7-days: rs = −0.21; all p’s > 0.05).

Significant correlations were found between the EMQ scores and recall of words on 7-day (rs = 0.33, p < 0.05), but not 2- (rs = 0.02) and 30-min (rs = 0.09; all ps > 0.05) delays and 7-day recognition accuracy (d′: rs = 0.44, p < 0.001); better everyday memory was associated with higher recall and better recognition accuracy.


This study found evidence of ALF in children with IGE. Moreover, children with more severe epilepsy had poorer long-term recognition of newly learned verbal materials. Memory problems displayed in day-to-day life were greater in IGE than in control children and were significantly associated with poorer recall and recognition after long, but not short, delays.

We found that children with IGE had difficulties with recall of information after long, but not short delays, which is typical of ALF. In our study (unlike in Davidson et al., 2007) ALF was not explained by reduced learning efficiency as (1) there was no between-group difference in the number of learning trials, (2) exclusion of participants who did not learn materials to criterion did not eliminate ALF, and (3) on the long delay the IGE participants had deficits not only in recall, but also in recognition of previously learned words. Taken together, our findings suggest that IGE participants have impaired long-term retention and impoverished long-term memory stores.

It is important to note that all IGE children in our study were still being treated for epilepsy and had experienced at least one overt seizure in the last 2 years, whereas 24% of children in the Davidson et al. (2007) study were in remission. Furthermore, we found that greater epilepsy severity was associated with worse long-term recognition. Therefore, the between study difference in epilepsy status may, in part, explain why the ALF was found only in our study. Nevertheless, whether seizures themselves (or a combination of epilepsy factors) compromised long-term memory formation is difficult to ascertain, as in our study seizure activity was not monitored during the retention period, and several studies of TLE patients (except Jokeit et al., 2001 and Mameniskiene et al., 2006) found comparable forgetting rates during retention periods, regardless of seizure activity (Bergin et al., 1995; Blake et al., 2000; Muhlert et al., 2011). In patients with IGE, ALF may also be secondary to the reduced capacity of neocortical storage sites (critical for long-term memory) due to diffuse neocortical abnormalities beyond the resolution of structural brain scans (Woermann et al., 1998). In addition, AEDs could have contributed to ALF, although most patients were taking monotherapy, which is less deleterious to cognitive functions than polytherapy (Jokeit et al., 2005).

In our study, parental ratings of children’s everyday memory were related to long-term (but not short-term) recall and recognition. This is consistent with previous adult studies that involved patients with left-hemisphere epilepsy focus (Blake et al., 2000) and patients with TEA (Butler et al., 2009). These findings are not surprising, as in everyday life, information frequently needs to be retained and recalled after long periods of time (i.e., days or months, rather than minutes). This pattern also testifies to the ecological validity of paradigms that require recall of information after long delays, and suggests that parental memory questionnaires may be used to screen for ALF.

Notably, prevalence of ADHD in our IGE sample was lower than in previous studies that involved children with epilepsy (Dunn et al., 2003; Dunn & Kronenberger, 2005; Hermann et al., 2007). This was possibly due to a reliance on preexisting diagnoses, rather than formal diagnostic assessments being conducted, which represents a limitation of our study. At present, however, there is no evidence that ADHD increases a risk of ALF.

In conclusion, this study provided evidence of ALF in patients with epilepsy who do not have a temporal lobe seizure focus/pathology: in children with IGE. Our findings are consistent with the consolidation theory, and suggest that factors other than a temporal lobe seizure site, such as epilepsy severity, may play significant roles in consolidation of memories into long-term stores. Clinically, it is important to note that (1) parental reports of everyday memory difficulties are related to long-term (but not short-term) memory recall, and (2) scores obtained on standardized memory tests that require recall of information after short delays do not account for deficits in long-term memory formation. Specifically, children with more severe epilepsy whose parents report higher levels of memory difficulties in day-to-day life are at risk of ALF, which can remain undetected on standardized testing. Given that standardized instruments do not capture this phenomenon, the development of appropriate clinical assessment tools is essential to improve diagnosis and develop treatments.


The authors thank the children and parents for their participation, Dr. Margaret Charles for her data analysis assistance, and acknowledge a University of Sydney scholarship awarded to Mr. Michael Gascoigne.


None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.