Contrary to previous literature (Mograss et al., 2006; Mograss et al., 2008) sleep neither modulated early or late ERP old/new effects in the first experiment. There was evidence of a late parietal old/new effect as well as an early frontal old/new effect in the item memory test in both groups in accordance with the assumption that successful performance in the item memory task is associated with both familiarity and recollection. Neither of these old/new effects was modulated by sleep, however. Comparable early mid-frontal old/new effects in both groups supports the view that item memory for which no contextual information is provided is not modulated by sleep (Drosopoulos et al., 2005). Concerning the late parietal old/new effect especially in the AM test, there are several possible reasons for not finding group differences. One possibility could be that the late parietal old/new effect is not sensitive enough to detect subtle changes in recollective processing which is actually supported by the results of the hits vs. incorrect answers comparison which takes the discrimination ability between old intact learnt stimuli and recombined ones into account and in which sleep effects were found in experiment one. Accordingly, it could be that beneficial effects of sleep in recognitionmemory studies could come about facilitated access to associative memories and the discrimination between old and rearranged word-pairs what is not necessarily reflected in amplitude differences in the late parietal old/new effect for whose estimation solely old and new items need to be contrasted. However, the late parietal old/new effect has been shown to be sensitive to the amount of information recollected, when manipulated experimentally within subjects (Vilberg et al., 2006; Wilding, 2000), and it has been shown to vary dependent on sleep vs. wake retention intervals before (Mograss et al., 2006; Mograss et al., 2008). Mograss and colleagues (2006; 2008) reported group differences in ERP old/new effects, next to general benefits in correctly classifying old stimuli after sleep compared to wake. For example, in their study from 2008, Mograss and colleagues showed frontal and parietal old/new effects to be more pronounced in a late time interval (555-765 ms) after sleep compared to wake.
The work presented in this thesis adds to the understanding of human memory, and especially of pattern completion and separation. In chapter 2, I have developed a novel recognitionmemory paradigm (named Memory Image Completion – MIC) particularly targeting pattern completion processes by manipulating stimulus completeness. Simultaneously, I have identified age-related recognitionmemory deficits suggesting a bias towards- but also a deficit in pattern completion. In chapter 3, I have replicated the findings of chapter 2, and eliminated perceptual confounds in memory performance using concurrent eye-tracking. The observed viewing patterns during encoding and retrieval could not account for the recognitionmemory differences across conditions and age groups, lending more validity to the task as a tool to assess pattern completion. In chapter 4, results of a collaborative case study have presented direct evidence that the hippocampus is differentially involved in pattern separation and completion. More precisely, a patient with selective bilateral DG lesions presented with memory performance indicative of deficient pattern separation, intact pattern completion, and a bias towards the latter. This was some of the first evidence directly linking the DG to pattern separation, while simultaneously excluding major contributions to pattern completion. Instead, a lesioned DG may send CA3 into overdrive promoting increased pattern completion. In chapter 5, I tried to further tackle hippocampal subfield involvement in the MIC, however, unfortunately, pattern similarity analyses remained inconclusive. Nevertheless, activity associated with pattern completion seemed to involve the STS, which indicates that a successfully retrieved pattern is reinstated there. Interestingly, some prominent age effects could be identified. Although the hippocampus was involved in more general retrieval in young but not older adults, overall older adults showed hyperactivity in the hippocampus and specifically CA3 suggesting that the hippocampal neural circuit does change with age. Additionally, generally reduced PhC-activity alongside a specific reduction during novelty processing revealed another affected site in aging. In chapter 6, a new segmentation protocol was developed in cooperation with other groups to enable accurate analyses on MTL regions including PhC, PrC, ErC and all hippocampal subfields, because the neuroanatomical literature has advanced in recent years and it is important to ensure that MR research is based on the appropriate anatomy.
was also used to study the involvement of specific brain areas in the formation of social recognitionmemory after an initial social encounter, mimicking the learning session in a social memory test. Male mice showed increased c- Fos synthesis in the MeA, the medial preoptic area and the piriform cortex, whereas the number of c-Fos-positive cells in the dHC area was not significantly affected (Ferguson et al. 2001, Richter et al. 2005b, Engelmann 2009, Samuelsen and Meredith 2011). The discrepancies between the findings on c-Fos activation in the dHC of mice after social stimulation, and the effect of HC lesions, challenges the interpretation of data from IEG activation in the context of memory formation and illustrates the need to do more accurate quantifications. In this context, it has been demonstrated that distinct parts of the HC are involved in different behaviours. This functional dissociation is supported by its anatomical connectivity and gene expression, and therefore a more detailed look at IEG synthesis by analysing each of these HC parts might help to clarify its involvement. However, instead of dividing the HC by anatomical connectivity according to its septotemporal axis, recent studies analysed the involvement of specific HC areas such as the CA2 in social recognitionmemory. CA2 projects almost strictly inside the HC itself, to CA1 and CA2 (contralateral) and CA3 (Hitti and Siegelbaum 2014), and back projects to the medial entorhinal cortex layer II (Rowland et al. 2013), from which it also receives olfactory responsive inputs (Gnatkovsky et al. 2004, Boisselier et al. 2014). Recent studies suggested CA2 to be critical for an intact social recognitionmemory (Hitti and Siegelbaum 2014), since inactivation of CA2 pyramidal cells or lesions in this region impaired social recognitionmemory without impacting other forms of HC-dependent memory (Stevenson and
and especially the “Remember/Know” procedure were severely tackled by Hum- phreys (e.g., Humphreys, Dennis, Chalmers, & Finnigan, 2000) and Dunn (2004), Yonelinas et al. adjusted the “Remember/Know” procedure in order to remove a confound in previous imaging and ERP studies, namely that regions or waveforms associated with recollection may have not been related to recollection per se but to the higher level of confidence associated with it. In this procedure, subjects gave either a "Remember" judgement or – if they could not recollect anything – a 5-scaled confidence rating. Thus, recollection-related regions were identified as those showing larger activity for "Remember" responses compared with confi- dently recognised but unrecollected items, whereas familiarity-related regions were those in which activation covaried with confidence reports. Regions associ- ated with recollection included parahippocampal gyri and the hippocampal forma- tion, whereas familiarity was associated with precuneus activation. There were also distinct patterns of frontal activation, but importantly, regions showed virtually no overlap (there was also no familiarity-related activity near the perirhinal cortex, see Henson et al., 2003). However, Dunn and Dennis (submitted) argued that un- der a GM perspective, all 5 response categories of Yonelinas et al. can be ar- ranged on a single continuum. They modelled the data using differently shaped activation functions obtained within a signal detection framework, showing that all major results of Yonelinas et al. could be modelled within a GM account. In par- ticular, they argued that the fact that Yonelinas et al. had found no region more active for high confidence familiarity judgements as compared to "Remember" responses was actually inconsistent with the assumption of two independent proc- esses. Marginally, it should be noted that their account of Yonelinas et al.'s hippo- campal activation data in terms of a functional association of hippocampal activity and decisional factors seems somewhat far-fetched. Similarly, Wais, Wixted, Hop- kins, and Squire (2006) used receiver operating characteristics in a study on young adults, patients with limited hippocampal lesions, and matched controls, and argued that the hippocampus supports both recollection and familiarity com- ponents of recognitionmemory (see also Squire & Zola, 1998).
Previous clinical research found that invasive vagus nerve stimulation (VNS) enhanced word recognitionmemory in epileptic patients, an effect assumed to be related to the activation of brainstem arousal systems. In this study, we applied non-invasive transcutaneous auricular VNS (tVNS) to replicate and extend the previous work. Using a single-blind, randomized, between-subject design, 60 healthy volunteers received active or sham stimulation during a lexical decision task, in which emotional and neutral stimuli were classified as words or non-words. In a subsequent recognitionmemory task (1 day after stimulation), participants’ memory performance on these words and their subjective memory confidence were tested. Salivary alpha-amylase (sAA) levels, a putative indirect measure of central noradrenergic activation, were also measured before and after stimulation. During encoding, pleasant words were more accurately detected than neutral and unpleasant words. However, no tVNS effects were observed on task performance or on overall sAA level changes. tVNS also did not modulate overall recognitionmemory, which was particularly enhanced for pleasant emotional words. However, when hit rates were split based on confidence ratings reflecting familiarity- and recollection-based memory, higher recollection-based memory performance (irrespective of emotional category) was observed during active stimulation than during sham stimulation. To summarize, we replicated prior findings of enhanced processing and memory for emotional (pleasant) words. Whereas tVNS showed no effects on word processing, subtle effects on recollection-based memory performance emerged, which may indicate that tVNS facilitates hippocampus-mediated consolidation processes.
Using ERPs to tap into familiarity and recollection, Rhodes and Donaldson (2007b) measured contribution of these processes to associative recognition of three word-pair types: lexical com- pounds (e.g., mars – bar), semantically and associatively related pairs (e.g., lemon – orange) and semantically related but unassociated pairs (e.g., violin – guitar). At test all three types of rela- tions elicited an equally large late parietal old-new effect. Conversely, the early frontal old/new effect, was present only for compounds that is in the condition in which two elements formed pre-experimentally unitized representations. A number of other studies (Diana, Yonelinas, & Ran- ganath, 2008; Quamme et al., 2007; Bader et al., 2010; J¨ager et al., 2006; Yonelinas et al., 1999; Diana et al., 2010) also delivered a considerable support for the unitization hypothesis by showing that pre-experimentally unrelated items, when bound together during the encoding and learned as a unitized whole, can be recognized on the basis of familiarity. Quamme et al. (2007) offered par- ticipants to encode unrelated word pairs by embedding them as separate members into a sentence frame or by forming a compound from the novel pair. A group of participants was then given an associative recognitionmemory test during which they were instructed to respond on the basis of their feeling of familiarity for the word pair. Another group of participants performed a standard associative recognition test (Experiment 2). Quamme et al. observed that under familiarity-only instructions participants performed significantly better on word pairs encoded in the compound condition than in the sentence condition, whilst no such differences were obtained under stan- dard recognition instructions. In another experiment (Experiment 1) Quamme et al. demonstrated that amnestic patients in spite of their general impraiment on associative recognition tests per- form relatively well when stimuli are unitized into novel compounds during the encoding (see also Giovanello et al. (2006) for a similar effect of pre-experimentally existing compounds).
At the onset, neuro-cortical areas of interest were mainly expected in the medial temporal lobe. However, as observed in other studies using the subsequent memory effect for verbal material, no medial temporal lobe activity was found during encoding (Henson et al., 2005). During retrieval, medial temporal lobe activity was identified for recognizable targets compared to targets forgotten throughout – left parahippocampal cortex activity in the different context condition and left hippocampus and right parahippocampal cortex activity in the same context condition. Analysis of these results demonstrated two properties of the medial temporal lobe. First, the medial temporal lobe engages sub-regions depending on the nature of the retrieval cue. While a novel context-target cue elicited activity in the left parahippocampal cortex only, presentation of the encoded context-target ensemble induced right parahippocampal cortex and left hippocampal activity. Second, activity in the medial temporal lobe does not predict the recognition failure of recognizable items effect, but indicates successful retrieval of encoded memory information during both recognition tests. Instead of medial temporal lobe areas, another neo-cortical area showed robust context- dependent activity in a wide variety of contrasts during the study and test phase of the experiment – the bilateral anterior insular cortex. After an extensive review of insular literature, in accordance with other cognitive studies, the thesis proposes three functions sub-served by the anterior insular cortex within cortical connectivity networks. In the present study, all of these functions are demonstrated in one memory paradigm.
The examination of non-target ERP old/new effects revealed evidence for developmental changes across all age groups in the neural correlates of source memory. A parietal non-target old/new effect, the ERP correlate of strategic recollection, was obtained for adolescents and adults but not for children. This pattern closely parallels the age differences in source memory performance observed in this study. By this, the present study replicates previous findings that the ability to strategically recollect source information is less matured in pre-adolescent children (Czernochowski et al., 2009, 2005). Most notably however, the present results, based on the combined analysis of changes in behavioural performance and neural activity, extend previous findings as they suggest that strategic retrieval processes greatly improve in late childhood and emerge with adolescence. In this way, the approach followed here has revealed a close correspondence between functional changes in source memory and those which have been suggested to occur in other domains of cognitive control in this age range (Paus, 2005). This suggestion is attested to by studies which have demonstrated that the transition from childhood to adolescence is marked by strong improvements in inhibitory control in the oculomotor domain (Munoz, Broughton, Goldring, & Armstrong, 1998; Williams, Ponesse, Schachar, Logan, & Tannock, 1999).
Depending on the information available for classifier design, one can distinguish between supervised and unsupervised pattern recognition ([Therrien, 1989, p. 2], [Jain, 1986, p. 6-7]). In the first case there exists a set of labelled objects with a known class membership. A part of this set is extracted and used to derive a classifier. These objects build the training set. The remaining objects, whose correct class assignments are also known, are referred to as the test set and used to validate the classifier's performance. Based on the test results, suitable modifications of the classifier's parameters can be carried out. Thus, the goal of supervised learning, also called classification, is to find the underlying structure in the training set and to learn a set of rules that allows the classification of new objects into one of the existing classes. The problem of unsupervised pattern recognition, also called clustering, arises if cluster memberships of available objects, and perhaps even the number of clusters, are unknown. In such cases, a classifier is designed based on similar properties of objects: objects belonging to the same cluster should be as similar as possible (homogeneity within clusters) and objects belonging to different clusters should be clearly distinguishable (heterogeneity between clusters). The notion of similarity is either prescribed by a classification algorithm, or has to be defined depending on the application. If objects are real-valued feature vectors, then the Euclidean distance between feature vectors is usually used as a measure of dissimilarity of objects. Hence, the goal of clustering is to partition a given set of objects into clusters, or groups, which possesses properties of homogeneity and heterogeneity. It is obvious that unsupervised learning of the classifier is much more difficult than supervised learning, nevertheless, efficient algorithms in this area do exist.
equality built into the principle of desert. The “achievement principle” is rather based on social rights and does not suspend them, although there is a necessary tension between those forms of recognition. If someone is in a good position to experience social esteem for his or her achievements, which allows him or her to be independent of protection from social rights, for example because of a high income and high social status, then this person has a recognition-based interest to opt against such a welfare provision altogether, because this would increase his or her position and therefore his or her social esteem in relation to those who are dependent on those welfare rights. So, even if, from a recognition-theoretical perspective that takes into account the society as a whole, meritocracy has to be off-set by the protection provided by social rights, equality of opportunity and the fair distribution of social esteem, the individuals or groups that compete do not have such motivation. This theoretical assumption also allows for in-depth empirical social research into distributional conflicts as struggles for recognition, in which certain groups have good reasons to opt against welfare while others have good reasons to opt for it. This is also reflected in the actual formation of capitalistic welfare states, where some tend more towards the “achievement principle” from which those who benefit are already better off – this can be called the Matthew effect (Wade 2004) – whilst others tend to provide more inclusive support and also to interfere with market results (Heath 2011).
This level of categorization, term ed the basic level (Rosch, Mervis, Gray, Johnson, and Boyes- Braem, 1976) or the entry level (Jolicoeur, Gluck, and Kosslyn, 1984), is the level at which objects are most quickly and easily categorized. Subordi nate-level classification (e.g., the chair in the upper left corner is a kitchen chair) typically takes some what longer, and superordinate-level categoriza tion (e.g., th a t’s a piece of furniture) takes even longer (Jolicoeur et al., 1984). In the context of a psychophysical experim ent, the term recognition sometimes means something other than entry- level categorization. In some experiments subjects must decide w hether test objects were seen pre viously in the experim ent (e.g., old-new recogni tion), or subjects must decide whether two images depict the same object (e.g., same-different judg ments or match-to-sample judgments). Although conceptually these tasks are not equivalent to en try-level categorization, they tell us a great deal about the features that are processed by the visual system. Clearly, successful entry-level categoriza tion is not the end of visual processing. It would not go unnoticed, for example, if someone were to break into your house and replace each piece of furniture in the den with a different piece having the same name. Likewise, we have no difficulty distinguishing the different objects in Fig. 2, al though each would be immediately classified as a chair. Visual processing and visual memory go much deeper than entry-level categorization. These other tasks help us to understand visual pro cessing at these deeper levels.
Many different experiments were performed to evaluate our system. Note that as a part of pre-processing step, eye extraction was applied on the images and the images were normalized to 56×46 size. We carried out experiments on two independent and different databases, one is the ORL and the other is the xm2vts. Both sets include a number of images for each person, with variations in pose, expression and lighting. The ORL set includes 400 images of 40 different individuals where each individual is represented by 10. The system was trained using 5 images for each person from this set. For the xm2vts set, we have used 1180 images for 295 different individuals with each individual represented by 4 different images. In this case, 2 images of each individual were used for training. The xm2vts database images are taken at different sessions (different days). The experiments on this database test the robustness of the proposed system under the variation in time conditions of the images. Different timing means different hair style, different clothes and different “moods”. Fig. 3 shows examples of the xm2vts database used for this experiment. The recognition rate was 93.5% for this experiment, while under the same conditions the MPEG-7 face recognition method achieved 89.5%.
But I want to stick to the point, which experiences of non- or misrecognition constitute an injustice and in which relation this stands to other forms of injustice. Ingram mentions (Ingram 2018 , 73) that not all persons deserve recognition; non- or misrecognition is therefore justiﬁed at least when persons refuse legitimate recognition to others, as in the case of racists. That is true. But three observations are central here: First, it is not only about recognition as an interpersonal relation, but also about the institutionalization (and materialization) of recognition or its negative counterparts. The behaviour of a racist is immoral, but for a social theory and critical theory of justice, the case of institutionalized racism is particularly relevant. For example, in the case of exclusive social and educational policies or global development policies that reproduce colonial logics. In the institutional structure, however, it is more di ﬃcult to determine, who deserves recognition or misrecognition, because people are always actors in di ﬀerent, overlapping social ﬁelds with their own antagonistic rules and forms of recognition. Social rights and market esteem do not go hand in hand.
For DNA, CpG motifs in eukaryotic DNA have been identified (Bird 1986; Krieg et al. 1995). For rRNA, the major constituent of total cellular RNA, known modifications include modified nucleotides such as conversion of uridine to pseudouridine, methylation of 2`- hydroxyls (2`-O-methylation) and methylation of bases at different positions (Maden 1990; Decatur and Fournier 2003). However, the overall frequency of 2`-O-methylated bases is rather low and the distribution is not random (Maden 1990). These nucleic acid modifications of DNA and RNA represent a structural feature, which enables the immune system to distinguish self from non-self-nucleic acids (Krieg 2002; Diebold et al. 2004; Heil et al. 2004; Ganguly et al. 2009). However, it has recently become clear that this discrimination is primarily achieved by the intracellular localization of the TLRs, which allows the recognition of viral nucleic acids released into endosomal compartments (Barton et al. 2006). In contrast, self-nucleic acids are rapidly degraded in the extracellular environment and fail to access endosomal compartments (Barton et al. 2006). A possible exception to this situation may be the case of apoptotic cells, which might protect RNA from extracellular degradation and deliver it to the endosomes of phagocytes (Diebold et al. 2006).
Hence, in order to be able to accumulate knowledge (i.e., to constantly encode new information but simultaneously retain old memories), a post-learning process of memory stabiliza- tion or consolidation is required. While the acquisition and retrieval of new information occur mainly and most efficiently during wakefulness, there is ample evidence that memory consolidation (i.e., reactivation and redistribution of newly encoded memory from the temporary into the long-term store) primarily and most efficiently takes place during sleep, a pe- riod when information flow into the brain is strongly reduced (for review, see [ 55 ]). Newly acquired information is initially stored within a memory buffer, the hippocampus. During sleep, these memories become repeatedly reactivated and thereby gradually transferred into long-lasting memory net- works. Specific networks are located within the neocortex, with the temporal dynamics of these reactivation phenomena being orchestrated by a fine-tuned interplay between fast field oscillations originating in hippocampal layers (i.e., sharp wave-ripple complexes [ 56 , 57 ]), thalamo-cortically generat- ed sleep spindles [ 45 , 58 ], and cortical SOs (~ 0.75 Hz [ 59 ]). In recent decades, the vast number of studies supporting the beneficial role of sleep in memory has inspired several theo- ries about the underlying mechanisms of sleep-associated memory consolidation. In early sleep and memory research, a main focus relied on “macroscopic” estimates of sleep, that is, the amount of NREM or REM sleep. Earlier experimental evidence suggested that NREM sleep and REM sleep differ- entially modulate the consolidation of declarative and non- declarative memories, respectively (i.e., the dual process
these asymmetric recalls as a self-deception strategy motivated by self-image concerns. This finding is consistent with previous theoretical and empirical studies on motivated memory revealing an asymmetric recall of feedback depending on whether individuals receive good or bad news about their relative performance (Bénabou and Tirole, 2002; Gottlieb, 2014; Li, 2017; Chew et al., 2018; Zimmermann, 2018). More generally, it contributes to the literature showing that individuals have motivated cognitive limitations even in the absence of risk and uncertainty (Exley and Kessler, 2018), selective memory being one of these self-serving biases. We complement the previous studies on motivated memory by showing that individuals also use selective memory in social interactions (the only previous evidence came from Li, 2013) and by revealing the crucial role of personal responsibility in this process. Indeed, the asymmetric recalls that we identified are no longer observed when decisions are made at random by the program. Moreover, our study shows that incentivizing correct recalls increases the percentage of dictators’ correct recalls when they chose the altruistic option but has no effect when they chose the selfish option. This suggests that when dictators are given a monetary incentive to provide a memory effort, they allocate this effort to retrieve the memory of desirable rather than undesirable information in terms of image. Like Zimmermann (2018), we interpret the fact that incentives generate more accurate recalls as evidence against complete forgetting. Individuals selectively suppress bad news (in the case of Zimmermann, 2018) or selectively retrieve good news (in our case).
breaks and/or regime switching (e.g., Diebold and Inoue (2001), Perron (1989), Per- ron and Qu (2007), Davidson and Sibbertsen (2005), Granger and Ding (1996)) unit roots (e.g., Hall (1978), Nelson and Plosser (1982), Perron (1988), Phillips (1987)), learning dynamics (e.g., Alfarano and Lux (2005), Chevillon and Mavroeidis (2011)), nonlinearity (e.g., Chen, Hansen, and Carrasco (2010), Miller and Park (2010)), as well as other mechanisms (e.g., Parke (1999), Calvet and Fisher (2002)). While these approaches all identify plausible mechanisms generating a long memory behavior, the search for a simple structural explanation for long memory is still actively ongoing (especially for the popular “fractionally integrated” processes). The goal of this paper is to identify a new, diﬀerent and arguably more universal mechanism.