Session 6: Thursday 11 February, 3pm-4pm

We had participants collect audio, accelerometry, GPS and emotion data continually over a two week period. A week later, we presented four alternative forced choice trials in which they had to identify where they had been at given times. To model the data, we created a memory network for each individual which contained nodes representing GPS, accelerometry and audio clusters as well as time (hour, day, week) and discrete emotions. Links were added between nodes that co-occurred most frequently with a constraint on the maximum degree. Each trial was modelled by adding the outer product of the time cue vector with itself to the memory network (simulating short term plasticity) and then cycling activations to find the primary eigenvector. The dynamic eigen network model with max degree learning rule captured 51.3% of participant choices (well above chance) exemplifying generalization at retrieval (c.f. Hintzmann, 1988) through the dynamic reconfiguration of the eigenstructure of the memory network.

When asked to recall autobiographical memories, people of independent cultures tend to recollect specific memories that focus on individual experiences, whereas people from interdependent cultures tend to recollect general and routine events that focus on collective activities and social relationships. Developed using research from independent cultures, the Self Memory System (SMS) model of autobiographical memory proposes that the recollection of specific autobiographical memories requires access to the hierarchically organized autobiographical knowledge base via either direct (the memory comes to mind immediately) or generative (memories from each level of the knowledge base are sequentially sampled before the final memory is reported) retrieval. Under this model, memories may be classified as belonging to a Lifetime Period (LTP; top of the hierarchy), being a General Event (GE), or containing Event Specific Knowledge (ESK; bottom of the hierarchy). The primary aim of the present study was to investigate cultural differences in retrieval of autobiographical memories to determine whether the SMS model can be generalised to interdependent cultures. Additionally, we aimed to explore cultural differences in the types (general or specific) and details (internal or external; following the Autobiographical Interview Schedule scoring manual) of the retrieved memories. Participants (Mean Age = 23.15 years, SD = 6.93) were recruited from The University of Newcastle, Australia (N=93) and The University of Dhaka, Bangladesh (N=70). They were presented with 12 cue words (4 each of positive, negative, and neutral) and asked to recall an autobiographical memory to each cue. After each recollection, participants were asked whether the memory came to mind immediately (direct retrieval) or they used other memories to help them retrieve the memory (generative retrieval). Memories were first coded for specificity (specific or general). Specific autobiographical memories were then coded based on the SMS model to determine the autobiographical knowledge level and whether the access sequence followed the anticipated LTP>GE>ESK. Memories were also coded for internal (episodic) versus external (semantic) details. In contrast to previous findings of cultural differences in the proportion of specific versus general memories retrieved in response to a cue word, no main effect of culture was observed. However, when cue valence was considered, cultural differences emerged. Compared to Australians, Bangladeshi participants recollected significantly fewer specific autobiographical memories in response to both positive and negative cues, but there was no difference for neutral cues. The proportion of specific autobiographical memories that were accessed following the SMS strategic search (i.e., LTP>GE>ESK) was higher for Australians than Bangladeshi participants. Although there was no overall difference in the number of details recollected, there were cultural differences in the type of details retrieved. Specifically, Australians recollected a higher number of episodic (i.e., internal) details and a lower number of semantic (i.e., external) details than Bangladeshi participants. Implications of these findings, both the similarities and differences across independent and interdependent cultures, for the SMS model of autobiographical memory will be discussed.

We develop a critical test of the two-high-threshold model of memory. Based on the approach taken by Kellen and Klauer (2014, Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 1795-1804), we focus on a ranking task in which one target and n-1 non-targets or foils are ranked in decreasing order of perceived likelihood of being the target. The resulting distribution of interest is the probability that the target is ranked in position k. We show that the two-high-threshold model predicts that the hazard function of this distribution increases monotonically for k > 1. We generalize this result to the case in which each foil has a potentially different probability of being detected under the model. This in turn led us to an investigation of the Poisson binomial distribution, showing that it too has a monotonically increasing hazard. Because the Poisson binomial distribution may have many other applications outside of memory research and outside of psychology, we believe this result to be both novel and significant.

The applied field of eyewitness memory has recently introduced Signal Detection Theory (SDT) models for the purpose of measuring identification procedure performance. While these models assume eyewitnesses have access to continuously graded memory strength, the field has historically maintained assumptions more aligned with discrete-state models, such as the two-high-threshold (2HT) model. To test these two opposing assumptions, we collected ranking data from 2,045 participants across two online between-subjects eyewitness identification experiments. Memory strength was manipulated over three levels (strong to weak) and the resulting hazard functions were calculated. Using a critical test of threshold representation (Chechile, R. & Dunn, J., 2021, A critical test of the two-high-threshold model of the recognition memory ranking task: Analysis of hazard functions [Paper presentation]. Australasian Mathematical Psychology Conference, Newcastle) the data shows that eyewitnesses do not simply 'guess' their responses when completing a lineup identification task. We conclude that SDT models remain the most appropriate set of models for the analysis of eyewitness identification data.

Psychological research is full of the longitudinal data in which a researcher is to compare two or more groups (experimental and control) over the time (pretest, posttest and follow-up). Mixed analysis of Variance is the most commonly used statistical method in such situation. This method can not meet a main assumption in which all correlations between each pair of observations within subjects must be same. An alternative method is GEE in which distribution assumption is robust, estimate for standard errors is more precise, and various covariance matrices can be chosen. To do this, using the data obtained from 40 depressed male patients (20 patients for Cognitive Behavioral Therapy and 20 patients for waiting list control group) in 3 time points (pretest, posttest and follow-up) were analyzed using Mixed ANOVA and GEE with SPSS-26 at the significance level of 0.05. The results indicated that GEE is more accurate than Mixed ANOVA (GEE-R squared =0.48 and Mixed ANOVA R squared=0.31). In this situation, GEE is more accurate and efficient than Mixed ANOVA.

The DRM paradigm is widely used in false recognition research. In the task, participants study lists of thematically related words and often identify unstudied related items as being ‘old’; these responses are presumed to reflect basic memory processes. Across two experiments, we find evidence that people sometimes relied on non-memory/strategic responses based on the relatedness of the test cue when they believed it to be of discriminative value. In Exp 1, our modelling indicated that the subjective experience of true memory partially differs from that associated with responses towards unstudied related words. In Exp 2, false recognition increased when participants were given explicit knowledge of the thematic label for thematically ambiguous study lists. We discuss the natural role of reasoning in memory, and strategies to identify/reduce its influence in false recognition tasks to isolate the pure products of memory.

The free association task provides a glimpse into the organizational structure of concepts in memory, and has been used by theorists as a benchmark for computational models of semantic processing. The most successful match to free association data comes from a high-level probabilistic approach called Topics modeling (Griffiths, Steyvers, & Tennenbaum, 2007). Topics models treat concepts as probability distributions over a set of discrete themes, or topics. To date, there has been no successful process-model of the free-association data-set. One path forward is to use high-dimensional vector representations as items in a broader control architecture, however, such approach decouples the encoding process from retrieval. We offer an alternative approach using a Dynamic Eigen Net (DEN), an associative net with dynamic eigenvectors, as a way to simultaneously model encoding and retrieval. We start with a replication of the Topics model of free-association, ensuring our corpus is matched to the corpus they used in their simulations. We then train a DEN, along with three other popular semantic representation algorithms. Our results show that a DEN provides a similar match to data as to the Topics model, while providing a better match than the three other algorithms. We suggest DENs as possible process-model compliments to the probabilistic account.