Measurement models for visual working memory—A factorial model comparison.

Abstract

Several measurement models have been proposed for data from the continuous-reproduction paradigm for studying visual working memory: The original mixture model (Zhang & Luck, 2008) and its extension (Bays, Catalao, & Husain, 2009); the interference measurement model (Oberauer, Stoneking, Wabersich, & Lin, 2017), and the target confusability competition model (Schurgin, Wixted, & Brady, 2020). This article describes a space of possible measurement models in which all existing models can be placed. The space is defined by three dimensions: (1) The choice of a activation function (von-Mises or Laplace), the choice of a response-selection function (variants of Luce's choice rule or of signal detection theory), and whether or not memory precision is assumed to be a constant over manipulations affecting memory. A factorial combination of these three variables generates all possible models in the model space. Fitting all models to eight data sets revealed a new model as empirically most adequate, which combines a von-Mises activation function with a signaldetection response-selection rule. The precision parameter can be treated as a constant across many experimental manipulations, though it might vary with manipulations not yet explored. All modelling code and the raw data modelled are available on the OSF: osf.io/zwprv Visual WM Measurement Models 3 Working memory for simple visual features is often studied with a continuous-reproduction task, in which participants are asked to remember an array of visual objects characterized by a simple visual feature varying on a continuous dimension, such as color or orientation. At test, one array item is selected at random as the target, and participants try to reproduce the target's feature on a continuous, usually circular, response scale, such as selecting its color on a color wheel.Research using this task has benefited substantially from measurement models, spearheaded by the model of Zhang and Luck (2008). They model the distribution of responses as a mixture of two components: A circular normal distribution (i.e., a von-Mises distribution) centered on the target's true feature, and a uniform distribution on the circle. This model -which I will refer to as the mixture model -is motivated by the assumption that WM has a limited number of slots, each of which can represent one array item with a certain precision. When the target is represented in a slot, responses are modelled by the von-Mises distribution. When the target is not maintained in a slot, the person has no information about it and must guess, leading to a uniform distribution of responses. The model has two free parameters, the precision of the von-Mises distribution, reflecting the precision of features stored in a slot, and the mixture weight of the von-Mises distribution, reflecting the probability that an array item is stored in a slot. The mixture model has soon been extended by Bays et al. (2009) by a third component of the mixture, reflecting responses centered on non-target features in the array. The exte...

Perceptual similarity judgments do not predict the distribution of errors in working memory

“…1). Furthermore, our findings on the advantage of the von Mises over the exponential responsiveness function are consistent with recent results reported by Oberauer (2021). By combining features of different descriptive models of WM within a space of possible models and fitting them to analogue report WM data, Oberauer demonstrated that a version of a model with a von…”

An interference model for visual and verbal working memory

“…Whereas in the first variant, the amount of noise (relative to the strength of the memory signal, as represented by Ac and Aa) is represented by the weight parameter b, in the second variant, it is represented by σ. A recent comparison of measurement models for the continuous-reproduction task (Oberauer, 2021) revealed that models using Luce's choice rule on exponentially transformed memory signals as in Equation ( 6) fit the data better than models with Luce's choice rule on untransformed signals (Equation ( 5)).…”

A tutorial for estimating mixture models for visual working memory tasks in brms: Introducing the Bayesian Measurement Modeling (bmm) package for R

“…Additionally, the fact that the IMM assumes continuous activations underlying the retrieval of memory representations adds a third account to the explanation of capacity limits in visual working memory, that is capacity being limited by bindings (Oberauer, 2021).…”