Accelerated long-term forgetting can become apparent within 3-8 hours of wakefulness in patients with transient epileptic amnesia.

Stimuli

Four categorical word lists were designed entitled “Animals,” “City,” “Nature,” and “Groceries.” Each list consisted of 16 category-related words, and all lists were matched for British word frequency (spoken and written) using the British National Corpus (BNC) database (BNC website: http://www.natcorp.ox.ac.uk/corpus/index.xml ), and for the psycholinguistic measures familiarity, imaginability, concreteness, and number of letters using the Medical Research Council (MRC) Psycholinguistic database (MRC website: http://websites.psychology.uwa.edu.au/school/MRCDatabase/uwa_mrc.htm ). Pilot work confirmed that the four categorical lists were matched for word-list learning over two trials and for recall performance 30 min and 1 week after acquisition.

Delayed recall

Participants’ retention of the learned word-list material was tested after four intervals: 30 min, 3 hr, 8 hr, and 24 hr after word-list learning (see Figure 1 ). To minimize potential masking of ALF by repeated testing of the same list, participants had to recall a different word list at each test interval. At all test intervals, participants were tested in the same location in the same room as during word-list learning.

Lists were probed in the counterbalanced order in which they had been presented during list learning. Prior to recall, the title of the to-be-remembered categorical word list (Animals, City, Nature, or Groceries) was presented on the laptop screen as a recall cue. As at immediate recall, participants were free to recall the words in any order, and they were asked to indicate when they had completed their recall.

Throughout the 30-min delay subsequent to word-list learning, participants were presented with a set of 300 complex, everyday-life photos, which they were asked to look at carefully and try to remember for an unrelated memory test. This task formed part of a different study and acted merely as a filler task in the present experiment. The experimenter remained with the participants until the 3-hr delay interval tests were completed. During the 3–8-hr interval, the experimenter was sometimes present, sometimes not, according to the convenience of participants. There was no systematic difference between testing of TEA patients and controls.

One week after word-list learning, participants were asked to recall all four lists again using the categorical titles as recall cues. Thereafter, a yes/no word-recognition test was conducted. In this test, the 16 original words (i.e., the targets) of each list were intermixed with 16 related foils. Thus in total, the recognition test consisted of 64 target items and 64 foils. Within each category, we manipulated the relatedness between targets and foils, such that six foils were semantically distant—phonetically related; five foils were semantically close—phonetically unrelated and five foils were semantically distant—phonetically unrelated. For example, in the case of the Animals categorical word list, “mouse” was used as a foil for “moose” (semantically distant—phonetically related), “lion” was used as a foil for “tiger” (semantically close—phonetically unrelated), and “snake” was used as a foil for “bear” (semantically distant—phonetically unrelated).

As during word learning, words were presented visually on a laptop screen. For each word, participants had to indicate verbally whether or not they had been presented with that word 1 week earlier.

Participants were not informed about the delayed memory tests. However, after completion of the 8-hr interval testing session, they were explicitly asked not to think about any of the tests they had undertaken during that day.

Delayed recall

We computed percentage retention scores for each participant for each delay interval. Percentage retention scores control for potential individual and group differences as well as any interlist variation at immediate recall. The latter was important because a different category list was probed at each delay interval. We calculated the percentage retention scores by dividing the number of words recalled after a given delay by the number of words recalled for that specific word list at learning Trial 2, and multiplying the quotient by 100 [(recall score after delay/recall score for that specific list at learning Trial 2) × 100]. Because all four lists were probed after 1 week and at learning Trial 2, we computed an average list-retention score for each participant for the 1-week word-recall test.

Word-recognition test

Calculations of d ’ values were made from hit and false-alarm rates: z (hit rate) – z (false-alarm rate). Hit rate refers to the number of original stimuli identified as “Old”, divided by the total number of original stimuli presented in the recognition test. False-alarm rate refers to the number of foils identified as “Old”, divided by the total number of foils presented in the recognition test.

Forgetting rates

Forgetting rates were list-specific in that a forgetting score was computed for each learned list from learning Trial 2 to one of the 4 early delay intervals (early forgetting), and from that early delay interval to the 1-week delay interval (late forgetting). The forgetting rates over the early (≤24 hr) and late (1 week) delay intervals were calculated as

Early forgetting = (learning Trial 2 recall score – recall score at specific delay interval)/(learning Trial 2 recall score) × 100.

Late forgetting = (recall score at specific delay interval – 1-week recall score)/(recall score at specific delay interval) × 100.

Guess correction

The use of categorical word lists and cued-recall tests can raise the tendency for participants to guess (see, e.g., Huff, Meade, & Hutchison, 2011 ; Tulving & Pearlstone, 1966 ). Therefore, in addition to analyzing raw retention scores, we also analyzed “guess-corrected” scores. We did so by applying a guess correction to word-list recall during learning and delayed recall, followed by computation of guess-corrected percentage retention scores. Word recall was corrected by dividing the number of correct words recalled by the total number of category-related words recalled, and multiplying this quotient by the number of correct words recalled: correct recalls × correct recalls/(correct recalls + incorrect recalls). For example, if 8 out of the 10 recalled category words were included in the presented categorical word list, the guess-correction factor would be 8/(8 + 2) = 0.8, resulting in a corrected recall score of 8 × 0.8 = 6.4.

Statistical Analyses

We applied a combination of independent t tests and mixed-factors ANOVAs to examine memory scores across the delay intervals in the two groups. Planned comparisons were carried out between pairs of delay intervals (30 min, 3 hr, 8 hr, and 24 hr) to examine changes in retention over these intervals. We applied Pearson correlations to examine associations between forgetting rates over the early (≤24-hr) and late (1-week) delay intervals, as reported previously ( Butler et al., 2012 ; Hoefeijzers et al., 2013 ; Muhlert et al., 2010 ; Wilkinson et al., 2012 ). Such correlations can provide insight into whether ALF is associated with an early or a later memory deficit. We used Pearson correlations to examine the relationship between “NART-predicted verbal IQ” and 1-week retention scores. ANCOVAs with covariate NART-predicted verbal IQ were run to examine whether the reported memory findings persisted when controlling for the subtle group difference in NART-predicted verbal IQ.

The Greenhouse–Geisser correction for nonsphericity was applied if the sphericity assumption (according to the Mauchly’s test of sphericity) was violated. Effect sizes for the ANOVAs were determined using partial η 2 , where 0.14 is a large effect ( Stevens, 2002 ). The α level was set to 0.05 for all analyses.

One-Week Performance

Figure 2 shows average percentage word-list retention and word recognition 1 week after word-list learning. Percentage retention after 1 week was significantly lower in the TEA patients than in the controls, t (25) = −6.357, p < .001, r = .79. Moreover, performance on the 1-week recognition test, that is, d ′ score, was also significantly lower in the TEA patients than in the controls, t (25) = −3.087, p < .01, r = .53 (see online Supplementary Data 2 for hit and false-alarm rates). The number of words recalled after 1 week (i.e., absolute scores) correlated significantly with the 1-week word-recognition performance in the TEA patients ( r = .608, p < .05) and in the controls ( r = .676, p < .01).