Qualitative aspects of learning, recall, and recognition in dementia

Objective: To determine whether learning and serial position effect (SPE) differs qualitatively and quantitatively among different types of dementia and between dementia patients and controls; we also wished to find out whether interference affects it. Materials and Methods: We administered the Malayalam version of the Rey Auditory Verbal Learning Test (RAVLT) to 30 cognitively unimpaired controls and 80 dementia patients [30 with Alzheimer’s disease (AD), 30 with vascular dementia (VaD), and 20 with frontotemporal dementia (FTD)] with mild severity on the Clinical Dementia Rating Scale. Results: All groups were comparable on education and age, except the FTD group, who were younger. Qualitatively, the learning pattern and SPE (with primacy and recency being superior to intermediate) was retained in the AD, VaD, and control groups. On SPE in free recall, recency was superior to intermediate in the FTD group ( P < 0.01 using Bonferroni correction). On recognition, the AD and VaD groups had more misses ( P < 0.01), while the FTD group had more false positives ( P < 0.01). Conclusion: Quantitative learning is affected by dementia. The pattern of qualitative learning remains unaltered in dementia in the early stages.


Introduction
Learning is a process by which information is acquired; it forms the background against which information is encoded and stored in memory. The process of learning and recall involves a defi nite patt ern. The study of this patt ern in the healthy and the diseased can enhance our understanding of the organization of memory in the brain and the eff ect of disease on it.
The Rey Auditory Verbal Learning Test (RAVLT) [1,2] is a widely used and well-validated word-list memory test that has been used extensively in assessing the learning curve, strength of memory aft er interference task, and patt ern of learning (serial positioning effects), as well as for measuring recognition memory. Normally, during free recall of a series of unrelated words from a word list, individuals tend to follow a patt ern known as the serial position eff ect (SPE), i.e., words from the beginning (primacy eff ect) and the end (recency eff ect) of the list are recalled bett er than the mid-list (intermediate) items. [3] There is considerable data available on the patt erns of learning in healthy individuals, and the patt ern of SPE has been shown to be relatively preserved in the healthy elderly. [4] However, litt le is known about what happens to the learning patt ern in patients with dementia. Pepin and Eslinger [5] in their study reported that patients with mild Alzheimer's disease (AD) demonstrated a 'U' shaped serial position curve similar to that the dementia group would diff er in their performance from the controls not only quantitatively, i.e., in the number of words recalled, but also qualitatively, i.e., in the patt ern of learning and recall.

Study design and participants
The design that we have employed in our study is a correlational one. The study patients were from the memory clinic of the hospital and included patients with AD (n = 30), VaD (n =30), and FTD (n =20), who had been diagnosed using the standard international criteria of National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA), [9] National Institute of Neurological Disorders and Stroke-Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS-AIREN), [10] and the consensus criteria, [11] respectively; for comparison we also had cognitively unimpaired healthy controls (n =30). The diagnosis and the dementia subtyping of patients att ending the memory clinic was fi nalized in a consensus conference, involving the neurologists, psychiatrists, and neuropsychologists, and using the clinical, biochemical, neuroimaging, and neuropsychological data. The consensus conference was blinded to the RAVLT data of the patients participating in the study. The cognitively unimpaired participants were selected from the community aft er a clinical evaluation and screening using the Addenbrooke's Cognitive Examination, [12,13] Instrumental Activities of Daily Living in Elderly, [14] Hospital Anxiety and Depression Scale, [15] and Clinical Dementia Rating Scale. [16] The AD and VaD groups were comparable with the controls on age. The FTD group, however, as expected, was younger. All the four groups were comparable on education [ Table 1]. All patients had disease of mild severity on the Clinical Dementia Rating Scale (CDR~1). All participants were native speakers of Malayalam, a Dravidian language of South India. All participants gave informed voluntary consent to participate in the study, which had received the approval of the Institute Ethics Committ ee and was carried out in compliance with the regulations of our institution.

Tools and administration
The Malayalam version of the RAVLT was administered to all participants by a trained neuropsychologist, a native speaker of Malayalam. The items in the original RAVLT were translated into Malayalam by a linguist profi cient in both English and Malayalam. To validate the interpretation it was back-translated by another person who was also profi cient in both languages. Due to the lack of a Malayalam equivalent for the word 'ranger,' the word 'carpenter' was used in place of the word 'ranger.' The word 'carpenter' has a Malayalam equivalent and has high imageability. Because no published word-frequency data was available for the Malayalam words, we chose words based on colloquial frequency. For the recognition task, along with the words in list A and B, phonemic and semantic distracters were prepared.
The participants were read out the list of words (list A) at a pace of one word per second. The scores obtained on trials 1 to 5 were used as a measure of learning and the scores obtained on the fi ft h trial as a measure of the pre-interference immediate recall. The distracter list (list B) was then read out and the participant was asked to recall it. Following this, the participant was asked to recall the words from list A; this was used as a measure of post-interference free recall. The delayed recall task was administered aft er 20 min and was followed by the recognition task in which the examiner read aloud a list of 50 words (this list included words from both list A and B and words phonemically or semantically related to them) from which the participants had been instructed to identify the words in list A.

Scoring
The total number of correct responses in each trial and the order of word recall in the fi rst trial, immediate pre-interference, immediate post-interference, and delayed recall trials were recorded on the scoring sheet. On the recognition task, 'hits' refer to the number of words correctly identifi ed from list A, and 'misses' refer to the words from list A that were not identifi ed. The words incorrectly identifi ed as present in list A constitute the false positives.
To study the SPE, we looked into the patt ern of recall for all the 15 items in the list. Further, we divided the entire list of 15 words into fi ve parts of three words each-the fi rst three words representing primacy, the middle three words representing intermediate, and the last three words representing recency. We used the scores on trial 1 to study SPE in free recall.

Statistics
ANOVA was used to compare the means across the diff erent groups. Post hoc analysis was done using the Bonferroni test. The paired t-test was used for intra-group comparison of performance. Wherever multiple comparisons were done, a Bonferroni correction was applied for determining the signifi cance of the P value. A recognition discriminability index was calculated using the d-prime analysis.

Results
There was signifi cant diff erence (P < 0.01) between the normal controls and the dementia groups in the mean number of words recalled in each of the fi ve free recall trials, the immediate recall of words in the pre interference and post-interference trials, in delayed recall and in recognition trials [ Table 1]. Although the rate of learning was lower for dementia patients, they followed the same patt ern of learning as that of normal controls across the fi ve trials, i.e., improvement in the number of words recalled with each subsequent trial. When the three dementia groups were compared with one another, there was no signifi cant diff erence (P > 0.05) between the groups in any of the parameters of memory.
All the dementia groups performed bett er on the recognition task than on the delayed-recall task (P = 0.05). The d-prime value showed that controls demonstrated bett er recognition of words from list A than the dementia patients. The scores of the patient groups indicated that within the dementia groups, the VaD group performed bett er on recognition than the AD group, while the AD group performed bett er than the FTD group, although the diff erences were not statistically signifi cant.
With regard to the type of errors on recognition, the AD and VaD groups demonstrated more misses than the control and the FTD groups (P < 0.01). The FTD group demonstrated signifi cantly more false positives than the controls or the other two dementia groups (P < 0.01). Within-group comparison showed that the AD and VaD groups demonstrated signifi cantly more misses than false positives (P < 0.05). However, in the FTD group, the hits and false positives did not show a statistically signifi cant diff erence (P > 0.05). Figure 1 shows a decline in the mean words recalled immediately aft er interference in the controls and the patient groups. This decline in mean words recalled was signifi cant in all the three dementia groups as well as in the controls (P < 0.01) On the delayed recall, when compared to the immediate postinterference, recall of words was impaired in the patient group (P < 0.05) but not in the controls. Figure 2 demonstrates the phenomenon of SPE in free recall for all the 15 items in the list. The patt ern obtained for the patient group was similar to that for the control group. Figure 3 demonstrates the primacy/recency eff ect in all the four groups. From the graph it can be seen that primacy and recency were preserved in the AD and VaD groups, just as in the control group. However, the FTD group demonstrated only the recency eff ect. Statistical analysis showed that the AD and VAD groups demonstrated signifi cant primacy and recency eff ects, similar to that seen in the controls (P < 0.01 using Bonferroni correction). The FTD group showed a signifi cant recency eff ect compared to intermediate recall (P < 0.01 using Bonferroni correction).
We also studied the patt ern of SPE following interference to determine whether the patt ern of words recalled following    Test used-Multivariate ANOVA. In the recognition part of RAVLT, *FTD had signifi cantly more false positives than controls and **AD and VaD had signifi cantly more misses than controls interference would be similar to that in free recall. Following interference, the patt ern of the SPE was lost in dementia patients as well as in controls [ Figure 4].  In our study we found that dementia patients recalled fewer words than normal controls. Recall of fewer words in dementia patients is possibly att ributable to poor encoding of the stimuli presented to them. [17] Our study replicates the fi ndings of Bayley et al. [18] who used the California Verbal Learning Test (CVLT) and found that AD patients recalled signifi cantly fewer words than normal controls and had signifi cantly reduced primacy eff ect, with a relatively preserved recency eff ect.
In the dementia groups, we have seen that although the recall of words is less than in the controls, there is an improvement in scores with successive trials. Our fi ndings are consistent with the reports by Bigler et al. [19] The authors, based on their study of 94 AD patients on RAVLT, reported that AD patients had negligible improvement in the learning/retention curve with repeated trials. Burkart et al. [4] have also reported similar fi ndings based on their study of 44 AD patients and 24 nondemented controls. Becker et al. [17] studied 62 patients with mild AD and 64 elderly controls and found that AD patients did not have an abnormal rate of forgett ing; they concluded that poor initial encoding of the stimuli may be the cause of impaired recall in AD patients.   In the free recall and recognition task, the AD and VaD groups in our study were indistinguishable in the scores obtained.
The results are similar to that reported by Almkvist et al. [20] In addition, in our study, the mean recall by the FTD group in the pre-interference trial was similar to the recall by the AD and VaD groups.
On the post-interference and delayed recall trials, the three dementia groups remained comparable on the mean number of words recalled, suggesting that interference profoundly hampers the process of retrieval, independent of the type of dementia.
Nevertheless, in the patient groups there was a signifi cant improvement in scores on the recognition task when compared to that on the recall task, which suggests that some information is indeed still accessible for recognition though not for free recall. On the recognition task, the FTD group had more false positives than all other groups (P < 0.01). The reason is possibly att ributable to a greater tendency to perseverate with the 'yes' response. In contrast, the AD and VaD groups demonstrated more misses compared to the other two groups (P < 0.01).
The fact that these groups did not have more false positives indicates that their eff orts are not just random guesses but that recognition is indeed assisting the retrieval of information that is inaccessible for free recall. These results suggest that in the dementia groups encoded information is not completely lost (or disintegrated). Instead, either its tagging is lost or disrupted or, alternatively, the neural network(s) responsible for free recall is/are structurally disintegrated.
In this study, we have also focused on the qualitative aspect of the SPE phenomenon in demented and cognitively unimpaired individuals. In the cognitively unimpaired, the normal patt ern of SPE of recency being superior to primacy is well established in the works of early researchers like Nipher, [3] Deese and Kaufman, [21] Jahnke, [22] and Rundus. [23] The reason for a superior recency eff ect in these individuals has been reported by researchers like Raaĳ makers and Shiff rin, [24,25] and Gillund and Shiff rin. [26] In cognitively unimpaired individuals, items toward the end of the list are recalled bett er in free recall because the contents of short-term storage are available for free recall at the time of the test and also, perhaps, due to the retroactive interference during encoding, i.e., the earlier items suff er interference from the later ones in the list. [27] Our results in patients with dementia show that, as in controls, the SPE is preserved in free recall in the AD and VaD groups. However, none of the dementia groups showed a superior recency eff ect compared to the primacy eff ect. These fi ndings are consistent with that of Pepin and Eslinger [5] who also showed that in mild AD both primacy and recency were above the intermediate portion of the curve and the SPE is 'U' shaped as it is in normal individuals. They also report a fading of the primacy eff ect with increasing severity of dementia. However, the results in the AD group are in contrast to that found in other studies, which have indicated impaired primacy in AD patients. [6] In our study, the age of our patients could have contributed to the bett er primacy eff ect. The AD patients in our study are distinctly younger (68.93 ± 7.28 years) than the patients in the study by Buschke et al. (80.6 ± 6.4 years). Thus it is possible that with regard to SPE there may be an interaction between age and dementia.
Our results in the VaD group showed an SPE with superior primacy and recency eff ects compared to intermediate (P < 0.01). A similar study by Paul et al. [8] demonstrated an intact SPE for patients with mild VaD.
In our study, the FTD group demonstrated a superior recency eff ect compared to intermediate (P < 0.01), which could be accounted for by the fact that in FTD the central executive system in the frontal lobes may be functioning at slightly impaired levels. The three-component model of working memory by Baddeley and Hitch [28] explains the fi nding seen in FTD. As per the phonological loop storage system in their model, as the number of items in a list that need to be rehearsed increases, what happens is that before the fi rst item can be rehearsed it fades out of the memory storage. [29] According to Hashimoto et al., [30] to learn the words effi ciently, participants should inhibit the words already learnt, selectively att end to the unrecalled words, and actively rehearse them. The manipulation of the information and attention shifting is known to be the function of the central executive system. As this function may be impaired in FTD, it is possible that FTD patients are unable to inhibit the retroactive interference during serial learning trials.
Thus, the results of our study suggest that while, quantitatively, memory storage and retrieval is ravaged by dementing diseases, the organization of memory-encoding mechanisms in patients with dementia seems to be less affl icted by the disease in the early stages and remains largely the same as in cognitively unimpaired individuals. Qualitatively, the mechanisms of encoding are relatively preserved in mild dementia, though the mechanisms of free retrieval and cued retrieval are diff erentially impaired. In addition, between the controls and dementia groups, there are diff erences seen in the quantitative aspects of encoding and/or retrieval.
Based on the findings of this study, we conclude that quantitative learning is aff ected by dementia. However, the patt ern of qualitative learning, as measured by the SPE in free recall, remains largely unaltered by dementia in the early stages, suggesting that this type of learning is not aff ected by mild dementia.