Unifying network model links recency and central tendency biases in working memory

Abstract

The central tendency bias, or contraction bias, is a phenomenon where the judgment of the magnitude of items held in working memory appears to be biased towards the average of past observations. It is assumed to be an optimal strategy by the brain, and commonly thought of as an expression of the brain’s ability to learn the statistical structure of sensory input. On the other hand, recency biases such as serial dependence are also commonly observed, and are thought to reflect the content of working memory. Recent results from an auditory delayed comparison task in rats, suggest that both biases may be more related than previously thought: when the posterior parietal cortex (PPC) was silenced, both short-term and contraction biases were reduced. By proposing a model of the circuit that may be involved in generating the behavior, we show that a volatile working memory content susceptible to shifting to the past sensory experience – producing short-term sensory history biases – naturally leads to contraction bias. The errors, occurring at the level of individual trials, are sampled from the full distribution of the stimuli, and are not due to a gradual shift of the memory towards the sensory distribution’s mean. Our results are consistent with a broad set of behavioral findings and provide predictions of performance across different stimulus distributions and timings, delay intervals, as well as neuronal dynamics in putative working memory areas. Finally, we validate our model by performing a set of human psychophysics experiments of an auditory parametric working memory task.

Location- and feature-based selection histories make independent, qualitatively distinct contributions to urgent visuomotor performance

“…Indeed, a study that directly compared them found that their effects were uncorrelated, and concluded that they are mediated by separate mechanisms (Galluzzi et al, 2022). History effects are relevant not only to attention and visuomotor behaviors, as in the present study, but also to other cognitive functions such as working memory (Papadimitriou et al, 2015, 2017; Kuo, 2016; Akrami et al, 2018; Boboeva et al, 2023), value assessment (Failing and Theeuwes, 2018; Constantinople et al, 2019), or task switching (Wylie and Allport, 2000; Monsell, 2003; Stoet and Snyder, 2007). Given this, and the remarkable degree to which the performance of our monkey subjects was swayed by history-driven biases, we conclude that the contribution of recent history to behavioral variability is likely much larger than is generally appreciated, particularly under more naturalistic, less constrained conditions during which multiple such effects can be expressed.…”

“…When associated with threshold performance in perceptual tasks, such effects are often referred to as “serial dependencies” (Kiyonaga et al, 2017; Manassi et al, 2023; Cicchini et al, 2024). Many of these effects are specifically related to sensory representations kept in working memory, such as when a saccade is made to the location of a remembered visual cue (Papadimitriou et al, 2015, 2017), or during discrimination tasks in which two stimuli are presented sequentially, one after the other (Akrami et al, 2018; Boboeva et al, 2023). In other cases, the categorization of stimuli requires comparison with respect to an implicit boundary that must be represented internally, and the dependence on past trials is consistent with a boundary representation that is automatically updated as new stimuli are experienced (Ashourian and Loewenstein, 2011; Raviv et al, 2012; Hachen et al, 2018; Mendonça et al, 2020; Sheehan and Serences, 2022; Boboeva et al, 2023).…”

“…Many of these effects are specifically related to sensory representations kept in working memory, such as when a saccade is made to the location of a remembered visual cue (Papadimitriou et al, 2015, 2017), or during discrimination tasks in which two stimuli are presented sequentially, one after the other (Akrami et al, 2018; Boboeva et al, 2023). In other cases, the categorization of stimuli requires comparison with respect to an implicit boundary that must be represented internally, and the dependence on past trials is consistent with a boundary representation that is automatically updated as new stimuli are experienced (Ashourian and Loewenstein, 2011; Raviv et al, 2012; Hachen et al, 2018; Mendonça et al, 2020; Sheehan and Serences, 2022; Boboeva et al, 2023). In these types of study, the data can be generally understood as perception resulting from current sensory evidence being combined with or compared to prior information stored internally, and the observed biases are most obvious when the available sensory evidence is weak, unreliable, so prior information takes precedence.…”

Reduced perceptual error monitoring in autism

“…One indicates that ASD participants use recent prior information less than non-autistics in perceptual tasks. This is the case when inter-trial intervals are short and require fast trial-to-trial updating 1,53 or when external statistics are volatile such that optimal inference yields large trial-to-trial prediction updating. 5,64 The second indicates poor online error correction in motor tasks 7,8,65 , and perhaps reduced conflict resolution when fast responses are required.…”

“…We calculated the recency bias, namely the bias of the first tone towards the most recent trial's frequency. 1 The bias is the difference in success rate between bias+ trials (where this contraction produces a performance increase) and bias-trials (where performance decreases) 3,51 (but see 53 ). In line with previous work 1 , the bias for ASD participants was on average smaller than non-autistics (rank-sum test: z = 2.39, p = 0.017).…”