CEMSi Program

Please click on a presentation title to read its abstract and view its relevant slides (where applicable).

Week 1

Session 1

Zoom Link: https://unimelb.zoom.us/j/85915569495?pwd=MVpHdFZWR05JZEtmQzRsaDNtNDRBUT09

Melbourne Date/Time: November 23rd, 11:00 AM
Seoul Date/Time: November 23rd, 9:00 AM
Philadelphia Date/Time: November 22nd, 7:00 PM
St Louis Date/Time: November 22nd, 6:00 PM

Talk

  • Prof. Jeff Zacks- Event comprehension, prediction, and memory updating

    Memory for recent events allows humans and other animals to perform adaptively in new events, anticipating how activity will unfold based on previous experience. For example, during a first visit to a grocery store you might encode the location of the produce section, and this could guide your steps next time. However, if the store were rearranged, predicting where to go based on memory would lead to an error. Such memory-guided prediction errors can produce short-term costs but also can induce adaptive long-term memory updating. Converging evidence from eye movements, change detection judgments, and neuroimaging establish evidence for associative retrieval of previous event features, which leads to predictions about related new events. Results suggest that when such predictions lead to errors this induces memory updating. Young and older adults show both similarities and differences in behavioral and neural markers of such memory updating, pointing to potential targets for memory improvement.

    Click here to watch a recording of this talk.

Posters

  • Mary Vitello- When the wandering mind trips: event boundaries influence mind-wandering and memory for temporal structure

    Although we spend much our day attending to the external world, up to half of our waking hours may be consumed by internally-directed attention to our meandering thoughts (Smallwood & Schooler, 2004). By virtue of withdrawing our attention away from environmental stimuli, mind-wandering can disrupt our encoding of newly encountered information. It is less clear how mind-wandering affects our ability to segment our ongoing experiences into discrete events. Event boundaries provide a critical scaffolding for episodic memory that helps us organize the temporal flow of our experiences. In order to investigate how temporal memory is influenced by the ongoing fluctuations in attention during task performance, we adapted the Sustained Attention to Response Task (SART) paradigm (Robertson et al., 1997) in a manner that allowed us to index the momentary occurrence of mind-wandering as participants experienced a temporally structured task. The SART typically requires participants to respond to each of a continuous sequence of digit stimuli and to withhold response to an infrequently presented target; lapses of attention are associated with increased response time variability. In our 'Episodic SART' paradigm, participants viewed a series of images of common objects with colored borders and were tasked with responding to all images except those whose border color was a pre-specified target. Although the images themselves were never relevant to the task, the category of the images changed six times (e.g., toys, foods, clothing) as did the border color. These changes constituted event boundaries. Participants were unaware that they would later be tested on their memory for the temporal order and relative duration of the respective categories. As a validation of our paradigm's ability to index mind-wandering, we found increased behavioral response time variability preceding errors. Furthermore, we found a significant change in the slope of the variance time course (VTC) when comparing trials leading up to an event boundary to those following an event boundary, with the former having a positive slope (indicating increasing mind-wandering) and the latter having a negative slope (indicating boundary-induced diminution in mind-wandering). Critically, we found that the slope of the VTC following event boundaries predicted subsequent memory for the temporal order of events. These results suggest that we can track changes in attentional states as participants experience temporally structured events, and that even task-irrelevant category changes are processed as event boundaries and serve to momentarily reduce mind-wandering in a manner that is consequential for temporal memory.

    Click here to see slides.

  • Savannah Born- Does narrative network centrality predict both free recall and recognition memory?

    How is our memory of a movie or story affected by its narrative structure? It has been shown previously that two network measures of narrative structure, causal centrality and semantic centrality, predict later memory in a free recall test (Lee & Chen, 2021). The first measure, causal centrality, reflects the degree to which a given event is causally connected to other events in the story; this measure is derived from human judgments of which pairs of events in a movie or story have cause-effect relationships. The second measure, semantic centrality, reflects the degree to which a given event has shared elements or meaning with other events; this measure is based on event-pairwise similarity scores derived from a sentence embedding method (Universal Sentence Encoder). In this behavioral experiment, we tested whether recognition memory, like free recall, could be predicted by causal and semantic centrality. Participants (Memory condition, N=9) watched three separate short movies. The movies were 2:18 - 2:44 minutes in length, and were split into 10 to 15 events. Next, participants answered a series of multiple choice questions about each event in the movie (4-option recognition memory test). To validate the questions, a separate group of subjects will take the same test without seeing any of the movies (Naive condition). We compared test accuracy between our current sample (Memory) with that from the previously collected free recall test. Both were also compared to our two measures of network centrality. We predict that, across events, causal and semantic centrality will positively correlate with accuracy for both recognition memory and free recall, wherein memory performance is better for more central events. Our preliminary results suggest that performance in the Memory condition is correlated positively between recognition and recall tests for 2 out of 3 movies (r=.62, r=-0.30, r=0.44). In next steps, we will add additional movies to the sample, and examine how the relationship between recognition and recall performance is affected by restricting to the questions which are least biased, i.e., those for which Naive subjects perform at chance on the recognition memory test.

    Click here to see slides.

  • Dr. Maverick Smith- Older adults rely on prior knowledge to help encode and remember everyday activities

    How do we understand and remember everyday events? Does knowledge influence these processes, and do these processes change with age?

    Observers comprehend a viewed action by reconstructing it in a mental representation called an event model. The event model represents, among other things, the goals of those performing the action. It generates predictions for what will happen next, and it guides attention to goal-relevant information. Event models are updated, and event boundaries between events are perceived when predictions fail. Event boundaries are important because they often align with the beginnings and endings of goals. For instance, gardening is composed of several sub-events each defined by unique sub-goals, such as getting the supplies, pouring the potting soil, planting the plant, etc. Such activities are segmented at event boundaries into discrete events in episodic memory.

    Prior research found that age-related declines in remembering everyday information is partially explained by older adults' inability to segment activities effectively. However, semantic knowledge remains stable with age. Older adults may leverage their intact knowledge to offset episodic deficits in encoding and remembering new instances of familiar actions. Knowledge affects memory; however, little is known about how knowledge affects moment-to-moment encoding processes involved in constructing one's event model.

    We investigated if older adults use their prior knowledge when attending to, segmenting, and remembering everyday activities. Participants watched videos of activities that were familiar to older adults (balancing a checkbook, planting flowers), and activities that were familiar to young adults (installing a printer, setting-up a video game). Participants completed memory tests after watching each video.

    We found that older adults attended less to goal-relevant information, segmented activities less efficiently, and had poorer memory than young adults. More importantly, we also found that age-related deficits in event encoding and memory were reduced when older adults had prior knowledge of the activity, and that individual differences in encoding efficiency predicted subsequent event memory. These results suggest that older adults use knowledge to inform their event models when encoding new instances of familiar activities, and that encoding efficiency and memory suffers when older adults lack semantic knowledge.

    Results have implications for future interventions designed to improve memory in older adults. Interventions could be designed to help older adults use intact knowledge when encoding a new instance of an activity. Such an intervention should influence gaze patterns and event segmentation. Changes in attention and event segmentation should subsequently improve episodic memory.

    Click here to see slides.

Session 2

Zoom Link: https://unimelb.zoom.us/j/85118978851?pwd=VWRGSERqUUp0dDNaTTRBS1lPTUU2Zz09

Hyderabad Date/Time: November 24th, 7:30 PM
London Date/Time: November 24th, 2:00 PM
Philadelphia Date/Time: November 24th, 9:00 AM
St Louis Date/Time: November 24th, 8:00 AM

Talk

  • Prof. Klaus Oberauer-  Why is time good for working memory?

    The passage of time is commonly thought of as the scourge of memory, in particular short-term or working memory: Prominent theories assume that working memory representations decay over time. Researchers' focus on time-related forgetting has contributed to a neglect of the beneficial effects of time for short-term maintenance: Longer time between presentation of successive items of a memory set (i.e., inter-item time), usually improves performance. I will present a series of experiments testing several hypotheses for explaining that effect: (1) People use the free time for strengthening memory representations of already presented items, either through rehearsal, attentional refreshing, or elaboration. (2) Longer time enables better short-term consolidation. (3) Longer time between items increases their temporal distinctiveness. These hypotheses differ in what they predict for which items of a memory list benefit from longer time: (a) Rehearsal, refreshing, and elaboration predict the benefit to be retroactive (affecting already encoded items) and global (affecting all earlier items). Short-term consolidation predicts a local effect, benefitting only the most recently encoded item (retroactive), and perhaps the following one (proactive). Temporal distinctiveness predicts a predominantly local benefit (both retro- and proactive). Our results refute all these predictions: The beneficial effect of free inter-item time is global and proactive. The best available explanation for it so far is that encoding of an item depletes a limited resource, which is gradually replenished in the following inter-item interval (Popov & Reder, 2020), thereby increasing encoding resource for subsequent items.

    Click here to see slides.

    Click here to watch a recording of this talk.

Posters

  • Dr. Andrey Nikolaev- Neural correlates of episodic memory buildup in naturalistic viewing behavior

    In everyday life, we bind visual samples of the world into a coherent episodic memory via eye movements. Previous eye tracking research revealed that visual scrutiny, gaze transitions between event elements, and refixations during encoding predict successful memory formation. However, little is known about the neural mechanisms of these processes. Here, we combine EEG and eye tracking to investigate the buildup of episodic memory representations in naturalistic viewing behavior.

    Twenty-eight participants encoded a series of events. Each event consisted of elements from three categories (faces, places, objects) and was presented for 10 s. After a distractor task, memory was tested for all event-specific element combinations. To succeed in the test, participants had to separate the different events by making strong associations between within-event elements. The simultaneously recorded EEG and eye movements were analyzed at the encoding stage of the task.

    A major problem of EEG-eye movement coregistration in free viewing concerns the overlapping effects of sequential saccades on EEG. We overcame this by using the regression-based deconvolution method, which allows correction of such overlap as well as removal of the confounding effects from eye movement characteristics and ordering factors, such as fixation rank. After deconvolution, we extracted EEG theta and alpha activity in epochs from -200 to +400 ms relative to the fixation onset.

    As expected, subsequent memory performance for the whole event increased with the cumulative number of fixations within single elements and inter-category gaze transitions between elements. We found that these two gaze memory effects are associated with two distinct memory effects of the fixation-related EEG activity: theta synchronization over the frontal and centro-parietal areas, respectively. The frontal theta effect may indicate the sampling of individual elements, whereas the centro-parietal theta effect may reflect the binding of elements into a coherent episodic event. Furthermore, memory performance was predicted by gaze returns to already visited categories. This refixation effect was associated with a fixation-related alpha desynchronization over the occipital areas. This may indicate the pivotal role of refixations in temporal accumulation of visual information that is needed for building a coherent representation of the whole event.

    We conclude that the combination of theta synchronization and alpha desynchronization at the fixation level supports the buildup of a coherent episodic memory across sequential eye movements.

    Click here to see slides.

  • Nick Simonsen- Fractionating the reward-memory literature

    The topic of reward effects on memory has been rapidly increasing and is often viewed as a coherent literature. However, there are critical differences in the methodology underlying sets of studies, likely with different underlying memory mechanisms. Here we propose that two distinct procedures comprise the overall reward-memory literature: instructed and feedback. Instructed studies tell participants of item-value associations during encoding with rewards earned during memory retrieval. In contrast, feedback studies ask participants to make responses during encoding, with rewards provided as feedback; memory retrieval itself is unrewarded. Some feedback studies require participants to make responses related to the to-be-remembered items, while others require participants to respond to an initial prompt before presenting an unrelated stimulus. While both procedures involve feedback, the first set of studies involves item-related feedback, and the second set has item-unrelated feedback. Despite this characterization not encompassing all studies in the reward-memory literature, our fractionation provides a framework for studying and differentiating between different behavioural results and memory mechanisms.

    Click here to see slides.

  • Dr. Marie-Lucie Read- The Importance of the Fornix in Object-In-Sequence Memory

    Space and time are critical components of context for episodic memory, and the hippocampus is crucial to both types of processing. The fornix is the principal white matter tract connecting the hippocampus to extra-temporal lobe areas, and it has been evidenced to be an important contributor to successful learning and navigation in the spatial modality. However, its role in supporting temporal order information processing is unclear. Our study investigated whether inter-individual differences in microstructure measures of the fornix, parahippocampal cingulum and inferior longitudinal fasciculus, of young adults, were associated with inter-individual differences in performance of object-in-sequence retrieval. Diffusion tensor imaging and neurite orientation dispersion density imaging measures were acquired, and principal components analysis was used to reduce these measures into biologically interpretable components. The fornix microstructure component reflecting fibre coherence correlated with our object-in-sequence retrieval performance measure but not with a measure from the sequence learning phase. Conversely, microstructure measures of the parahippocampal cingulum and of the inferior longitudinal fasciculus did not correlate with object-in-sequence memory performance. Therefore, this study revealed the unique involvement of the fornix-pathway in aiding memory for objects with temporal context, that was not seen in the parahippocampal cingulum or inferior longitudinal fasciculus supported pathways.

    Click here to see slides.

Session 3

Zoom Link: https://unimelb.zoom.us/j/81274081253?pwd=VXJlb01xNkJlN01xNGJLUzRNcEtwUT09

Melbourne Date/Time: November 25th, 4:00 PM
Seoul Date/Time:
November 25th, 2:00 PM
Hyderabad Date/Time:
November 25th, 10:30 AM

Talks

  • Kevin Shabahang- Generalization over distinct associative substrates with a dynamic associative net

    Generalization is a defining feature of higher cognition and remains a challenge for extant models. We examine two qualitatively distinct associative substrates and show how an associative net we call Dynamic-Eigen-Net (Shabahang, Yim, Dennis, 2020) can provide a unifying account of generalization in both domains. One associative substrate encodes serial-order information, where generalization falls within the class of grammar induction problems, or the development of competence in the recognition and generation of serially ordered symbols (e.g., words) from a combinatorial domain (e.g., well-formed sentences). The other associative substrate encodes associations that are not organized according to any serial-order, where generalization requires exploiting structure involving more than a single degree of separation. Generalization in the latter case can be considered a superclass of property induction problems. Observing two associations, one between A and B (e.g., 'feather' and 'fly'), and another between B and C ('fly' and 'falcon'), should entail the presence of a third association between A and C ('feather' and 'falcon'). First we compare a Dynamic-Eigen-Net with a simple linear system and a commonly cited associative net, the Brain-State-in-a-Box (Anderson, Silverstein, Ritz & Jones, 1977). We show that Dynamic-Eigen-Net is more sensitive to the combinatorial structure of the input compared to the other two models. Then we compare another Dynamic-Eigen-Net with several semantic representation algorithms, such as Latent Semantic Analysis (Landauer & Dumais, 1997), and show that the Dynamic-Eigen-Net can match their performance. Our results demonstrate that the Dynamic-Eigen-Net can handle generalization within both associative substrates, providing a spreading activation account, where generic representations emerge as steady-states during retrieval.

    Click here to see slides.

    Click here to watch a recording of this talk.

  • Sophie O'Donnell- When did I post that? How we access memory-for-when on social media.

    Strategies for accessing temporal information of autobiographical memories may differ based on the qualities of the target memory. Two categories of models are proposed to account for day-to-day temporal memory access. Distance-based models propose dating occurs through accessing a decaying strength of a memory. Location-based theories propose performance is dictated by the availability of specific contextual information. Contextual source elements give clues to the temporal identity and their availability may bias strategy selection toward location-based memory strategies. Distance-based strategies are hypothesized to be utilised in situations where source elements are less available, and the required resolution of temporal information is low. They posit that accuracy declines with the age of the memory, rather than the presence of source elements.

    Experience sampling studies such as Dennis et al. (2017) demonstrate that location-based models can explain tasks where participants recall when they were at a specified location. The present study utilises participants' social media posts and emails as cues instead. These cues are less tied to location, routine, and therefore source element availability. It was hypothesised that this would bias participants toward distance-based strategies.

    Participants provided stimuli from Gmail, Twitter, Facebook, and Instagram from the 4 weeks immediately preceding their memory test. At test, they indicated the week, day-of-week, and time-of-day separately of each composition. Three forms of analyses showed results consistent with distance-based models. Firstly, accuracy was significantly lower with longer time delays for week judgments (t(34)=5.20, p< .001, d=1.30) and day-of-week judgments (t(34)=3.79, p= .001, d=0.63). There was no significant difference in accuracy for hour-of-day judgements with time delay (t(34)=-1.76, p= .087). Secondly, a confusion plot for week judgements shows errors appear to be distributed within temporal proximity to the correct response. Errors for weeks 1, 2, and 3 show a distribution significantly different from a random distribution that would be expected under location-based models (ps< .05). Thirdly, error scores compared responses to chance guessing with an empirical prior. A clear scale-decrement pattern was observed with accuracy decreasing with finer timescales . This is contrary to the findings of studies that use locations or media events as cues. The converging evidence of these analyses suggests that utilisation of cues that have less connection to physical places and events may allow observation of distance-based strategies.

    Click here to see slides.

    Click here to watch a recording of this talk.

Posters

  • Luz Viviana Sastre-Gomez- How long are real-life events?

    Research in event cognition has focused on how people perceive, remember and judge daily activities. Previous studies have found that in real-life events (Zhuang, Belkin, and Dennis, 2012; Sreekumar, Nielson, Smith, Dennis & Sederberg, 2018) participants generally segment between eight and sixteen events in a day and event durations follow an exponential distribution. However, there are still issues and questions such as, are the distributions for all event types the same? and is there a change in how people identify events and their duration according to the moment they are asked about the event? This research study aims to explore the temporal duration of self-reported events from daily life using contemporary sampling methods.

    Forty participants provided activity duration data as they went about their everyday lives. Participants selected their previous activity from a predetermined list (e.g. reading, working, drinking) and the time the activity started, seven times per day for 14 days. Descriptive analyses and activity duration modelling (a mixture model combining both a normal and an exponential process) were used to identify the distribution of event durations within activity types.

    Preliminary results show that most of the events present an exponential pattern of durations, while others show a bimodal pattern with a secondary frequency peak of around 120 minutes. Planned events such as meetings, working, exercising and transiting seem to show the secondary peak in duration around two hours, while, unplanned activities such as watching tv, reading, writing, eating, drinking, shopping, and using social media show a unimodal pattern.

    Although the preplanned events look like they have a characteristic time, they are still not a normal distribution; indeed, these events also have a substantial exponential component. In terms of duration extension in preplanning events, the results of this study are consistent with previous studies by Zhuang, et al (2012) and Sreekumar, et al (2018). These results might provide a better understanding of the mechanisms underlying the perceived temporal duration of real-life events because it shows that a lot of daily events do not have a characteristic duration.

    Click here to see slides.

  • Prajneya Kumar- Why do we remember some stimuli better than others?

    Recent work has shown that some face and scene images may have certain features that make them more memorable than other images. While similar investigations of word memorability do exist, a comprehensive analysis of the word features that explain memorability by using a pool of words that vary widely on many key dimensions such as concreteness, context variability, etc. is currently lacking. We analyze an existing word memorability dataset that uses cued-recall to probe people's memory to get preliminary insights into the word features that may predict memorability. Furthermore, to understand whether word memorability generalizes across languages and cultures, we translated the English word pool used in the previous memorability study to Hindi in order to conduct analogous cued-recall tests with Mechanical Turk participants to get independent measures of Hindi word memorability. Finally, we hypothesized that word memorability is related to the centrality of words in semantic networks and that the availability of multiple cues (i.e., semantic associates) in memory may facilitate the memorability of these central words. To this end, we will construct semantic networks independently for English and Hindi and present a preliminary analysis of the network properties of the high and low memorability words in both languages.

    Click here to see slides.

Week 2

Session 1

Zoom Link: https://unimelb.zoom.us/j/87464566464?pwd=TDlycllVZDRTaWtMVHdDVHhGSmdBZz09

Melbourne Date/Time: November 30th, 4:00 PM
Seoul Date/Time:
November 30th, 2:00 PM
Hyderabad Date/Time:
November 30th, 10:30 AM

Due to a cancellation, this session will be dedicated entirely to poster presentations.

Posters

  • Indupriya B.- Impairments in Valence Linked Retrieval of Episodic Autobiographical Memories.

    Objectives: Even though, Memory complaints are widely studied in patients with refractory epilepsies, structural and functional correlates of episodic autobiographic memories are poorly understood. In the present study we have explored the extent of the deficits in specific autobiographical memory generation in patients with post-surgery refractory epilepsies. As emotions are integral to autobiographical memory system we further explored the role of emotion valence (positive and negative) in epilepsy patients who underwent Amygdalohippocampactomy.

    Methods: Addenbrook Cognitive Examination (ACE-III), Digit Span test, Beck Depression inventory (BDI), and Autobiographical Memory test (AMT) were administered to post-surgery, seizure free epilepsy patients (n=10, 7M, 3F) and age-gender matched healthy controls.

    Results: Results of Mann Whitney U test and Wilcoxon sign rank test showed the presence of specific autobiographical memory deficit in refractory epilepsy patients who underwent Amygdalohippocampectomy (n=6,6M,p=.02) in comparison with  age and gender matched  healthy  controls.

    Conclusion: Exploring structural basis of valence linked impairments in generating specific autobiographical memories helps to understand Affect and Cognition interaction and subsequent impairments in patients with refractory epilepsies.

    Click here to see slides.

  • Jason Zhou- Similarity Affects Swap Error Probability in Source Memory

    Source memory is memory for the source or the origin of material stored in memory. Traditionally, models of source memory have attempted to differentiate between competing characterisations of source memory retrieval as a thresholded or continuous process using data from two-choice source tasks. However, the high degree of model mimicry in two-choice data has made differentiating between competing accounts difficult. More recently, research using continuous-outcome tasks, which yields more diagnostic data in the form of entire distributions of response precision rather than response proportions, has not considered the contribution of intrusions, or swap errors between items to these distributions of response outcomes, nor distributions of response times (RT) in addition to response outcomes. In this work, we introduce a circular diffusion account of source memory which addresses both of these shortcomings by modelling RT and error, as well as systematic errors due to intrusions among items presented in the same list.

  • Raina Zhang- Integrating Semantic and Orthographic Representations in Recognition Memory

    Global matching is a key concept in many models of recognition memory which posits that retrieval operates by matching the test probe against each of the representations stored in memory to produce a measure of global similarity. However, to date many models have not adopted principled representations of items. For word stimuli, while some memory models have adopted realistic semantic representations from semantic space models, little work has explored the consequence of integrating perceptual representation, and it remains unclear how representations on different dimensions are combined. The current study aimed to explore how semantic and orthographic representations are combined at memory retrieval. Specifically, whether they are combined multiplicatively, additively or through an additive combination followed by non-linear transformations. We varied the level of semantic and orthographic similarities between test probes and study content in a standard recognition memory paradigm using unrelated study lists. Semantic and Orthographic similarity for each test probe were combined via different methods to form a global similarity index which was regressed on the drift rate parameter of the Linear Ballistic Accumulator to jointly model recognition accuracy and latency within Hierarchical Bayesian frameworks. Models utilising different combining mechanisms will be compared and the best-fitting model will be determined.

  • Andy Sitoh- Using the Linear Ballistic Accumulator to Understand How List Length Affects Decision Making in Free Recall

    While the decline in recall performance with increasing list-length has been taken as evidence for inhibition between items in memory, previous investigations have largely looked into response probabilities or mean response latencies. In other words, they have not been extended to complete distribution of latencies. The present study aimed to provide a deeper understanding of how list-length affects free recall by jointly fitting both response probabilities and complete latency distributions from first recall data. The experiment was a delayed free recall task where five list-length conditions (ranging from 1 to 16 items) were manipulated within participants. The Linear Ballistic Accumulator model was subsequently applied to investigate what manipulations were required to fit the data. Specifically, a 2 x 2 x 2 factorial model comparison was conducted, varying in whether: (1) Memory-strengths inhibit one another; (2) across list-lengths, different levels of evidence were required to recall an item during the recall task; and (3) across list-lengths, different levels of evidence were already present before the recall task. There was no consistent support for the first parameterisation (memory-strengths inhibit one another), but there was consistent support for the other two parameterisations. These results challenge the assumption of inhibition between items in memory, though some caution must be taken as only first recalls were modelled. Instead, as list-length increases, greater levels of evidence were required for recall and greater levels of evidence were already present before the recall task. The study highlights the diagnostic value in jointly modelling both response probabilities and complete distributions of latency in differentiating mechanisms of recall.

    Click here to see slides.

Session 2

Zoom Link: https://unimelb.zoom.us/j/85702228608?pwd=c28wVjJkVktVbXpnS2NSNGc5OFppZz09

Melbourne Date/Time: December 1st, 11:00 AM
Seoul Date/Time: December 1st, 9:00 AM
Philadelphia Date/Time: November 30th, 7:00 PM
St Louis Date/Time: November 30th, 6:00 PM

Talk

  • Dr. Eitan Schechtman- The role of encoding context in memory reactivation during sleep

    Throughout our waking lives, new memories are formed continuously. Such memories never exist in a vacuum; they may be interlinked with other memories that share semantic features and with those that share temporal proximity. In this study, we explored the role encoding context plays in offline consolidation during sleep. Memory was tested using 72 images of single objects, each presented with a related sound. Human participants first formed idiosyncratic narratives linking sets of objects together. Then, they learned the distinct 2D spatial positions of each object. During non-rapid-eye-movement sleep, some object-specific sounds were unobtrusively presented, a technique termed targeted memory reactivation. This manipulation putatively reactivated object memories, thereby provoking consolidation. Results showed that reactivated memories benefited from this procedure. Crucially, memories that were contextually linked to those reactivated also benefited from the procedure. Moreover, multivariate pattern analyses revealed that context-specific electrophysiological patterns were reinstated following sound presentations during sleep. Overall, these results indicate that reactivation of memories during sleep is accompanied by the reactivation of features of encoding context and that this intervening processing impacts subsequent recall.

    Click here to watch a recording of this talk.

Posters

  • CANCELLED- Jordan Gunn- An Instance-Based Account of Context Maintenance and Retrieval

    This presentation has been cancelled due to unforeseen circumstances.

    In retrieved context accounts of memory search such as the Context Maintenance and Retrieval (CMR) model (Polyn & Kahana, 2009), representations of studied items in a free recall experiment are associated with states of an internal representation of context that changes slowly during the study period. Through these mechanisms the model accounts for organizational effects in recall sequences, such as the tendency for related items to be recalled successively. Specifications of the model tend to characterize these dynamics in terms of a simplified neural network, as building a single prototypical pattern of associations between each item and context (and vice versa) across experience. By contrast, models of categorization and other memory phenomena have increasingly converged on instance-based accounts (Hintzman, 1984) that conceptualize memory as a stack of trace vectors that each represent discrete experiences and support recall through parallel, nonlinear activation of traces based on similarity to a probe. To investigate the consequences of this distinction we present an instance-based specification of CMR that encodes associations between studied items and context by instantiating memory traces corresponding to each experience, and drives recall through context-based coactivation of those traces. We analyze the model's ability to account for traditional phenomena that have been used as support for the original prototypical specification of CMR, evaluate conditions under which the specifications might behave differently, and explore the model's capacity for integration with existing instance-based models to elucidate a broader collection of memory phenomena.

  • Ashwin Ramayya- Distributed brain networks for anticipatory modulation of human response times

    The brain uses learned (cognitive) expectations to guide even the simplest actions, but the specific neural circuits responsible for this link between cognition, which tends to rely on widespread neural processing, and motor control, which tends to rely on localized neural processing, is not understood. We used quantitative modeling and intracranial recordings from patients with medically refractory epilepsy to identify neural representations of expectation-dependent response-time (RT) variability. These representations, which encoded model-derived quantities that map onto certain motor-preparatory activity patterns and thus are often interpreted as having a localized implementation, were not restricted to motor regions but instead were evident in particular, widespread neural firing patterns with distinct timing, form, and anatomical source locations. Our results show that expectation-dependent modulations of action rely more on emergent network-wide dynamics than on local neural representations and illustrate the complex mapping between algorithmic and implementation-level descriptions of flexible information processing in the brain.

    Click here to see slides.

Session 3

Zoom Link: https://unimelb.zoom.us/j/88312705254?pwd=czhuNVlSR1plbEdWRGcwdHVPUlY4QT09

Hyderabad Date/Time: December 2nd, 7:30 PM
London Date/Time: December 2nd, 2:00 PM
Philadelphia Date/Time: December 2nd, 9:00 AM
St Louis Date/Time: December 2nd, 8:00 AM

Talk

  • Dr. Vencislav Popov- Intent Matters: Resolving the Intentional vs Incidental Learning Paradox in Episodic Long-term Memory

    Decades of research have established that the intent to remember information has no effect on episodic long-term memory. This claim, which is routinely taught in introductory cognitive psychology courses, is based entirely on pure-list between-subject designs in which memory performance is equal for intentional and incidental learning groups. In the current ten experiments, participants made semantic judgements about each word in a list but they had to remember only words presented in a specific color. We demonstrate that in such mixed-list designs there is a substantial difference between intentionally and incidentally learned items. The first four experiments showed that this finding is independent of the remember cue onset relative to the semantic judgement. The remaining six experiments tested alternative explanations as to why intent only matters in mixed-list designs but not in pure-list between-subject designs' inhibition of incidentally learned items, output interference, selective relational encoding, and a novel selective threshold-shifting account. We found substantial support for the threshold-shifting account according to which the intent to remember boosts item-context associations in both mixed- and pure-list designs; however, in pure-list between-subject designs, participants in the incidental learning group can use a lower retrieval threshold to compensate for the weaker memory traces. This led to more extra-list intrusions in incidental learning groups; incidental learning groups also showed a source memory deficit. We conclude that intent always matters for long-term learning, but that the effect is masked in traditional between-subject designs. Our results suggest that researchers need to rethink the role of intent in long-term memory.

    Click here to see slides.

    Click here to watch a recording of this talk.

Posters

  • Neomi Mizrachi- Motor-sensory dynamics during a free recall task

    Motor-sensory dynamics is an essential component of perception. During an episode, the scanning patterns of sensory organs affect the acquired information. Despite that, the inter-relations between motor-sensory dynamics and episodic memory are not often studied. To fill this gap, we designed a free recall task that took place in a virtual reality (VR) environment, in which participants could freely move. During the encoding stage participants were asked to explore a VR environment containing 15 everyday virtual objects. During the recall stage, participants were asked to recall the objects names in a different VR environment. Participants' eyes, hands and heads were tracked at both stages. During encoding, participants walked around, looked at (virtual) objects and sometimes moved objects (virtually). Our initial results suggest that two motor-sensory factors occurring during encoding could increase objects' recall probability: The action of moving an object and the total gazing time at an object. The order of objects' recall was correlated with the order of their locations along the walking trajectory during encoding but not with the order in which they were gazed at.

    Click here to see slides.

  • Dr. Nicola Sambuco- Emotional modulation in the default mode network during narrative imagery

    Authors: Nicola Sambuco, Margaret M. Bradley, Peter J. Lang

    The default mode network (DMN) is activated when constructing and imagining narrative events, with functional brain activity in one region, the medial-prefrontal cortex, hypothesized to be modulated during emotional processing by adding value (or pleasure) to the episodic representation. However, since enhanced reactivity during emotional, compared to neutral, content is a more frequent finding in both the brain and body in physiological, neural, and behavioral measures, the current study directly assesses the effects of pleasure and emotion during narrative imagery in the DMN by using a within-subject design to first identify the DMN during resting state and then assess activation during pleasant, neutral, or unpleasant imagery. Replicating previous findings, enhanced functional activity in the medial prefrontal cortex was found when imagining pleasant, compared to unpleasant, events. In addition, emotion-related activation was found when imagining either pleasant or unpleasant, compared to neutral, events, in other nodes of the DMN including the posterior cingulate cortex (PCC), angular gyrus, anterior hippocampus, lateral temporal cortex, temporal pole, dorsomedial prefrontal cortex (dmPFC), and ventrolateral prefrontal cortex (vlPFC). Pervasive emotional modulation in the DMN is consistent with the view that a primary function of event retrieval and construction is to remember, recreate, and imagine motivationally relevant events important for planning adaptive behavior.

    Click here to see slides.

  • Dr. Deborah Talmi- Using ERP measures of attention during encoding to constrain tests of the emotional Context Maintenance and Retrieval model of free recall

    Supported by vast behavioural and neural data, retrieved-context theory is the state-of-the-art in describing the rules which govern the chances that a present experience will come to mind later. It determines the likelihood of recalling an experience through the similarity between its encoding and retrieval "contexts" - abstract constructs implemented in brain states. In addition to similarity between temporally-contiguous brain states, retrieved-context theory recognises that semantic and experiential similarity also influence which experience would come to mind. eCMR (emotional Context-Maintenance and Retrieval) focuses on the emotional sub-dimension of experiential similarity as a way of explaining emotional memory enhancement in terms of more general memory processes.

    Click here to see slides.

Week 3

Session 1

Zoom Link: https://unimelb.zoom.us/j/86771071810?pwd=d1FhckRBc1hTalVwUHZHWWJYdGFyZz09

Hyderabad Date/Time: December 7th, 7:30 PM
London Date/Time: December 7th, 2:00 PM
Philadelphia Date/Time: December 7th, 9:00 AM
St Louis Date/Time: December 7th, 8:00 AM

There will be no posters for this session, as there will be two talks instead.

Talks

  • Prof. Geoff Ward- Towards an Integration Between Free Recall and Serial Recall: Examining a Free Recall Account of Transposition Error Gradients in Serial Recall

    In immediate serial recall (ISR), participants are presented with short sequences of items and must try to repeat them back in exactly the same order that they had been presented. Many theories of ISR assume that each item is associated with position information, whereas no such mechanism has been suggested for immediate free recall (IFR). One line of evidence supporting item-position associations is that there are error gradients in the distribution of order errors: i.e., when list items are recalled in the wrong output order, they tend to be output in neighbouring list positions, suggesting that each item is associated with more or less perfect order information. Reanalysing published ISR datasets from open and closed sets of items, the current work shows that if participants recall known runs of primacy and/or recency items of different lengths, then error gradients arise even when middle items (for which there is otherwise unknown order information) are distributed randomly between known sequences of primacy and recency items. The work suggests that error gradients in ISR can be obtained in the absence of serial order information other than the start and end sequences observed in more IFR-like outputs, representing a step toward theoretical integration of the two tasks.

    Click here to see slides.

    Click here to watch a recording of this talk.

  • Dr. Jacob Bellmund- Structuring time: The hippocampus constructs sequence memories that generalize temporal relations across experiences

    The hippocampal-entorhinal region supports memory for episodic details, such as temporal relations of sequential events, and mnemonic constructions combining experiences for inferential reasoning. However, it is unclear whether hippocampal event memories reflect temporal relations derived from mnemonic constructions, event order, or elapsing time, and whether these sequence representations generalize temporal relations across similar sequences. Here, participants mnemonically constructed times of events from multiple sequences using infrequent cues and their experience of passing time. After learning, event representations in the anterior hippocampus reflected temporal relations based on constructed times. Temporal relations were generalized across sequences, revealing distinct representational formats for events from the same or different sequences. Structural knowledge about time patterns, abstracted from different sequences, biased the construction of specific event times. These findings demonstrate that mnemonic construction and the generalization of relational knowledge combine in the hippocampus, consistent with the simulation of scenarios from episodic details and structural knowledge.

    Click here to see slides.

    Click here to watch a recording of this talk.

Session 2

Zoom Link: https://unimelb.zoom.us/j/84176127138?pwd=NVRJNG9wbGREZTFNTTFsY0ZPd2VmUT09

Melbourne Date/Time: December 8th, 4:00 PM
Seoul Date/Time:
December 8th, 2:00 PM
Hyderabad Date/Time:
December 8th, 10:30 AM

Talk

  • Prof. Muireann Irish- Temporal distance and event plausibility modulate semantic contributions to past and future event construction

    The capacity to envisage the future is a complex constructive endeavour that is typically held to rely upon integrity of the episodic memory system. While a robust literature has explored episodic contributions to future thinking, far less is known regarding the contribution of semantic memory. We have previously demonstrated that the ability to envisage specific and detail-rich events in the future is markedly compromised in semantic dementia; a neurodegenerative disorder characterised by the progressive deterioration of conceptual knowledge. Here, we build on these findings to explore how the degeneration of the conceptual knowledge base in semantic dementia impacts past and future thinking across different temporal contexts and levels of plausibility. A novel experimental task was used whereby participants recalled recent ("Last year") and remote ("10 years ago") past events and constructed near ("Next year") and far ("10 years' time") future events. An implausible condition further required participants to imagine spending a day on the moon. Relative to age-matched Controls, semantic dementia patients displayed striking impairments in remote memory retrieval with recent memory relatively preserved. Future simulation was globally disrupted in semantic dementia irrespective of temporal distance. Notably, semantic dementia patients displayed disproportionate impairments on the implausible condition ("day on the moon") relative to plausible event simulation. Voxel-based morphometry analyses based on structural brain scans revealed variable contributions of regions typically implicated in episodic memory (hippocampus, angular gyrus) and semantic memory (lateral temporal cortices) for past and future thinking, while novel event construction was predominantly associated with the hippocampus and ventromedial prefrontal cortex. We interpret these findings in relation to current theoretical frameworks emphasising the interplay between the episodic and semantic memory systems in the service of complex forms of event construction.

Posters

  • Yicong Zheng- The effect of working memory on testing effect: interaction between capacity and demands

    Ample evidence has shown that learned information that is practiced by testing outperforms that is practiced by additional study (i.e., the testing effect). However, it is unclear whether testing equivalently benefits everyone, such as people with different working memory capacities (WMC). Existing evidence showed inconsistent results regarding the moderating role of WMC on the benefits of testing. We hypothesize that WMC manifests influence on the testing effect when the working memory demands (WMD) exceed WMC. The current study adapted the paradigm used in Reder et al. (2016), which showed that item familiarity affects working memory demands. Subjects were trained to discriminate novel complex visual stimuli over three weeks, during which half of the characters appeared much more frequently. After familiarized with the stimuli, in week 4, subjects learned randomly formed associations between these stimuli and words. After initial encoding, half of the associations were practiced by cued-recall test without feedback, and the other half were restudied. The final test was given one day later, during which all associations were tested. The results showed a three-way interaction, such that both WMC and WMD moderate the testing effect. Specifically, while subjects with higher WMC showed the testing effect with both familiar and unfamiliar stumuli, the benefits of testing only emerged with familiar stimuli in subjects with lower WM. Moreover, the results showed a trend toward a revered testing effect for unfamiliar stumuli in subjects with lower WM; that is, restudied associations were remembered better than tested associations. The findings are consistent with the dual-process account of the testing effect, which states that subjects benefit from testing through a retrieval attempt process and a re-encoding process. The current results suggested that both processes deplete limited resources of working memory. The testing effect only emerges when the WMC is adequate for both processes. Implementing this interpretation, we adapted the Source of Activation Confusion (SAC) model to account for the current results.

    Click here to see slides.

  • Isabella Fonti- Understanding Human Memory for Where: Factors that Influence Free Recall of Everyday Events

    Free recall paradigms using word lists have enabled us to characterise different features of episodic memory, including serial position effects, list length effects and word frequency effects. Likewise, free recall tasks have contributed an understanding towards the organisational structure of episodic memory, including how individuals cluster responses based on temporal and spatial contexts. However, laboratory free recall tasks may not reflect the type of retrieval that people might engage in in everyday life. Employing experience-sampling methods, the present study aimed to understand whether these observed trends in free recall can translate to circumstances where individuals are required to remember locations they frequented on a certain day.

    Fifty-eight participants collected Google location data for 30 days and completed an online test where they selected markers on an interactive map if they believed it was a location they visited on that date. We observed that: (a) participants were more likely to recall where they frequented at the start and end of the day, showing primacy and recency effects, (b) there was a positive relationship between recall accuracy and number of locations visited, showing an inverse-list length effect, (c) successively recalled locations are likely to occur in close temporal sequence, showing temporal contiguity, but not asymmetry effects, (d) successively recalled locations are likely to occur in close spatial sequence, showing spatial clustering effects, and (e) frequently visited locations are more often recalled, however, frequently visited locations are also likely to act as prior-list intrusions. Overall, discriminability was greater for low frequency locations than for high frequency locations. In sum, we generalised some well-established free recall trends to Memory for Where. However, not all laboratory-based observations may extrapolate to real-world conditions.

    Click here to see slides.

Session 3

Zoom Link: https://unimelb.zoom.us/j/85773055461?pwd=endqOTVYOUtRYWVmU0lxajV5cU54UT09

Melbourne Date/Time: December 9th, 11:00 AM
Seoul Date/Time: December 9th, 9:00 AM
Philadelphia Date/Time: December 8th, 7:00 PM
St Louis Date/Time: December 8th, 6:00 PM

Talk

  • Prof. Florin Dolcos-  The Impact of Emotion on Memory for Context: Behavioral, Eye-Tracking, and Brain Imaging Evidence

    Emotional stimuli tend to capture attention, and hence are typically remembered better than neutral ones. However, not all aspects of emotional events are equally remembered, as unpleasant information tends to be remembered with fewer contextual details, possibly because of initial narrowing of attentional focus by emotional arousal. In extreme circumstances, this may lead to gist-based retrieval of traumatic memories, as observed in post-traumatic stress disorder (PTSD). First, I will introduce a model proposing that gist-based retrieval in PTSD is linked to "decontextualized encoding" due to extreme arousal, which in turn leads to non-specific retrieval of memories for distressing events and perpetuation of PTSD symptoms. Then, I will present novel behavioral, eye-tracking, and brain imaging evidence supporting the effectiveness of an emotion regulation (ER) strategy (focused attention, FA) in reducing both the emotional (re)experience and memory for emotional stimuli, during encoding and retrieval. Functional MRI data show that FA manipulation involves interplays among brain regions involved in cognitive control (prefrontal cortex), emotion processing (amygdala), and processing of contextual information (parahippocampal cortex). Interestingly, FA manipulation also leads to enhanced encoding of non-emotional contextual aspects of emotional stimuli, which holds promise for its usefulness as a training tool to "re-contextualize" memories and improve post-traumatic symptoms following distressing events. Finally, consistent with this idea, I will also present proof-of-concept evidence from an intervention study in student veterans showing that increased general ability to deal with emotional challenges following ER training is associated with neuroplasticity in the resting state functional connectivity of brain regions associated with cognitive and affective control.

    Click here to see slides.

Posters

  • Camille Gasser- Learned action sequences act as a scaffold for novel episodic memories

    As we navigate through the world, many of the actions we follow are highly familiar. We may take the same route to work, cook the same recipe for dinner, or run along the same path during a nightly run. However, layered upon these routine behaviors are novel episodic experiences: while on a run, for example, you might pass by an unfamiliar dog or listen to a new song. To date, a substantial body of research has explored the important question of how what we already know about the world (i.e., our prior knowledge) affects how we learn novel, related pieces of information. Nevertheless, few (if any) studies have directly examined the common situation we describe here: namely, how one's engagement in a well-learned sequence of actions might support their ability to remember new, unrelated information that is encountered in parallel. To investigate this question, we designed a novel behavioral paradigm in which participants ran errands through different stores. During each errand, they visited a sequence of aisles in the store (by pressing buttons on their keyboard) while collecting (i.e., viewing) trial-unique items. There were two store conditions: in the predictable store, the sequence of aisles that participants visited (and consequently, the sequence of behavioral responses they made) was always the same and was well-learned prior to their errands; in the random store, this aisle/response sequence changed with each errand. Across three separate experiments, we found that participants were better able to remember the order of novel items when they were encountered in the predictable versus the random store, suggesting that the presence of a well-learned action sequence during encoding might serve as a scaffold for temporal memory. Critically, no such benefits were observed for item or source memory, precluding the explanation that participants simply paid more attention to novel items when their environment was partially familiar. We interpret such results as consistent with the idea that engaging in a familiar action sequence during encoding might activate a stable, temporally-extended memory representation, which in turn allows novel pieces of information to be embedded within it. This link to a shared underlying structure then promotes memory for the temporal links between those items. Future work will use fMRI to examine how interactions between the hippocampus and mPFC might underlie this scaffolding effect.

    Click here to see slides.

  • Dr. Hongmi Lee- A generalized cortical activity pattern at internally-generated mental context boundaries during unguided narrative recall

    Current theory and empirical studies suggest that humans segment continuous experiences into events based on the mismatch between predicted and actual sensory inputs; detection of these "event boundaries" evokes transient neural responses. However, boundaries can also occur at transitions between internal mental states, without relevant external input changes. To what extent do such "internal boundaries" share neural response properties with externally-driven boundaries? We conducted an fMRI experiment where subjects watched a series of short movies and then verbally recalled the movies, unprompted, in the order of their choosing. During recall, transitions between movies thus constituted major boundaries between internal mental contexts, generated purely by subjects' unguided thoughts. Following the offset of each recalled movie, we observed stereotyped spatial activation patterns in the default mode network, especially the posterior medial cortex, consistent across different movie contents and even across the different tasks of movie watching and recall. Surprisingly, the between-movie boundary patterns were negatively correlated with patterns at boundaries between events within a movie. Thus, major transitions between mental contexts elicit neural phenomena shared across internal and external modes and distinct from within-context event boundary detection, potentially reflecting a cognitive state related to the flushing and reconfiguration of situation models.

    Click here to see slides.

  • Sam Audrain- Elaborating on the role of the hippocampus in constructing remote autobiographical memories

    The necessity of the hippocampus for the retrieval of remote autobiographical memories has been debated for decades. In relatively recent years, the retrieval of autobiographical memories has proven to be a temporally dynamic process, consisting of an early phase of memory construction and a later phase of detailed elaboration, involving the anterior and posterior hippocampi respectively. Interestingly, this dynamic retrieval process has not been investigated in the context of the hippocampal temporal gradient during retrieval of remote autobiographical memories.

    In the present study, we re-analyzed data from 40 participants who retrieved recent and remote memories in response to picture cues, and overtly recalled them in detail in the fMRI scanner. We previously reported a temporal gradient in the posterior hippocampus during what corresponded to the elaboration phase of autobiographical retrieval, with reduced posterior hippocampal activation at remote timepoints. These findings suggested the hippocampus was not active during the retrieval of remote memories. Here, we consider a previously unanalyzed earlier period of retrieval, where participants were searching for and selecting a memory (construction phase) to subsequently elaborate upon. We compared anterior and posterior hippocampal activation during autobiographical construction and during a control picture selection task for memories differing in remoteness.

    While we observed a temporal gradient in the posterior hippocampus during elaborative retrieval (reported in our previous work), we found no evidence of a temporal gradient during the earlier memory construction phase. Instead, we observed strong anterior hippocampus activity, regardless of remoteness of the memory. Our findings suggest that the anterior hippocampus can support autobiographical memory search and/or construction in perpetuity, and highlights that the phase of autobiographical retrieval analyzed as well as location along the long-axis of the hippocampus can lead to differing conclusions regarding the involvement of the hippocampus in retrieval of remote memories.

    Click here to see slides.