scholarly journals Representing uncertainty in the Rescorla-Wagner model: Blocking, the redundancy effect, and outcome base rate

2021 ◽  
pp. 14-21
Author(s):  
Stuart G. Spicer ◽  
Andy J. Wills ◽  
Peter M. Jones ◽  
Chris J. Mitchell ◽  
Lenard Dome
Keyword(s):  
Associative Learning ◽  
Associative Strength ◽  
Open Science ◽  
Base Rate ◽  
Cue Competition ◽  
Wagner Model ◽  
Modified Model ◽  
Redundancy Effect ◽  
Causal Status ◽  
Intermediate Value

It is generally assumed that the Rescorla and Wagner (1972) model adequately accommodates the full results of simple cue competition experiments in humans (e.g. Dickinson et al., 1984), while the Bush and Mosteller (1951) model cannot. We present simulations that demonstrate this assumption is wrong in at least some circumstances. The Rescorla-Wagner model, as usually applied, fits the full results of a simple forward cue-competition experiment no better than the Bush-Mosteller model. Additionally, we present a novel finding, where letting the associative strength of all cues start at an intermediate value (rather than zero), allows this modified model to provide a better account of the experimental data than the (equivalently modified) Bush-Mosteller model. This modification also allows the Rescorla-Wagner model to account for a redundancy effect experiment (Uengoer et al., 2013); something that the unmodified model is not able to do. Furthermore, the modified Rescorla-Wagner model can accommodate the effect of varying the proportion of trials on which the outcome occurs (i.e. the base rate) on the redundancy effect (Jones et al., 2019). Interestingly, the initial associative strength of cues varies in line with the outcome base rate. We propose that this modification provides a simple way of mathematically representing uncertainty about the causal status of novel cues within the confines of the Rescorla-Wagner model. The theoretical implications of this modification are discussed. We also briefly introduce free and open resources to support formal modelling in associative learning. Keywords: associative learning, prediction error, uncertainty, modelling, blocking, redundancy effect, open science.

scholarly journals Representing uncertainty in the Rescorla-Wagner model: blocking, the redundancy effect, and outcome base rate

2021 ◽  
Author(s):  
Stuart Spicer ◽  
Andy Wills ◽  
Peter M Jones ◽  
Chris Mitchell ◽  
Lenard Dome
Keyword(s):  
Associative Strength ◽  
Base Rate ◽  
Cue Competition ◽  
Wagner Model ◽  
Modified Model ◽  
Redundancy Effect ◽  
Causal Status ◽  
Intermediate Value ◽  
Effect Experiment ◽  
Better Than

It is generally assumed that the Rescorla and Wagner (1972) model adequately accommodates the full results of simple cue competition experiments in humans (e.g. Dickinson et al., 1984), while the Bush and Mosteller (1951) model cannot. We present simulations that demonstrate this assumption is wrong in at least some circumstances. The Rescorla-Wagner model, as usually applied, fits the full results of a simple forward cue-competition experiment no better than the Bush-Mosteller model. Additionally, we present a novel finding, where letting the associative strength of all cues start at an intermediate value (rather than zero), allows this modified model to provide a better account of the experimental data than the (equivalently modified) Bush-Mosteller model. This modification also allows the Rescorla-Wagner model to account for a redundancy effect experiment (Uengoer et al., 2013); something that the unmodified model is not able to do. Furthermore, the modified Rescorla-Wagner model can accommodate the effect of varying the proportion of trials on which the outcome occurs (i.e. the base rate) on the redundancy effect (Jones et al., 2019). Interestingly, the initial associative strength of cues varies in line with the outcome base rate. We propose that this modification provides a simple way of mathematically representing uncertainty about the causal status of novel cues within the confines of the Rescorla-Wagner model. The theoretical implications of this modification are discussed. We also briefly introduce free and open resources to support formal modelling in associative learning.

scholarly journals The redundancy effect is related to a lack of conditioned inhibition: Evidence from a task in which excitation and inhibition are symmetrical

2019 ◽  
Vol 73 (2) ◽  
pp. 260-278
Author(s):  
Tara Zaksaite ◽  
Peter M Jones
Keyword(s):  
Associative Strength ◽  
Individual Variation ◽  
Conditioned Inhibition ◽  
Causal Learning ◽  
Learning Effects ◽  
Inhibitory Effects ◽  
Wagner Model ◽  
Opposite Pattern ◽  
Redundancy Effect ◽  
Human Causal Learning

Rescorla and Wagner’s model of learning describes excitation and inhibition as symmetrical opposites. However, tasks used in human causal learning experiments, such as the allergist task, generally involve learning about cues leading to the presence or absence of the outcome, which may not reflect this assumption. This is important when considering learning effects which provide a challenge to this model, such as the redundancy effect. The redundancy effect describes higher causal ratings for the blocked cue X than for the uncorrelated cue Y in the design A+/AX+/BY+/CY–, the opposite pattern to that predicted by the Rescorla–Wagner model, which predicts higher associative strength for Y than for X. Crucially, this prediction depends on cue C gaining some inhibitory associative strength. In this article, we used a task in which cues could have independent inhibitory effects on the outcome, to investigate whether a lack of inhibition was related to the redundancy effect. In Experiment 1, inhibition for C was not detected in the allergist task, supporting this possibility. Three further experiments using the alternative task showed that a lack of inhibition was related to the redundancy effect: the redundancy effect was smaller when C was rated as inhibitory. Individual variation in the strength of inhibition for C also determined the size of the redundancy effect. Given that weak inhibition was detected in the alternative scenario but not in the allergist task, we recommend carefully choosing the type of task used to investigate associative learning phenomena, as it may influence results.

Cue Competition in Prospective Memory

1999 ◽  
Vol 85 (3) ◽  
pp. 1011-1024
Author(s):  
Amy M. Richards ◽  
E. Evan Krauter
Keyword(s):  
Associative Learning ◽  
Test Trial ◽  
Prospective Memory ◽  
Control Group ◽  
Cue Competition ◽  
Prior Training ◽  
Blocking Group ◽  
Cue Words ◽  
The Relationship

Prospective memory refers to remembering to perform a previously planned activity. Two experiments were conducted to see if effects of cue competition similar to blocking and overshadowing occur in prospective memory. Participants were led to believe that the experiments were about the relationship between memory and creativity. To test prospective memory, participants were instructed to mark cue words that would appear later in a task requiring the generation of sentences. In Exp. 1 ( N = 119) one group was told to place an “x” over the cue word “rake”; a second was told to mark two words of equal salience (“method” and “rake”); and a third group was told to mark two cue words of unequal salience (the highly salient word “monad” and “rake”). “Rake” was the only cue word that actually appeared in the task involving generation of sentences. Participants instructed to place an “x” over one cue marked the target cue “rake” more frequently than if told to mark two cues (an overshadowing-like effect). The frequency of marking “rake” was lowest on the first test trial if participants had been instructed to mark both “rake” and “monad.” In Exp. 2 (N = 43) a blocking group was trained to mark one cue word (“rake”) and a control group received no training. Two days later, all participants were instructed to mark two cues (“rake” and “method”) during a task involving the generation of sentences. Prior training interfered with performance to a new cue (“method”) given in combination with the pretrained cue (“rake,” a blocking-like effect). These experiments demonstrate the existence of cue competition in prospective memory and suggest the possibility of applying theories of elementary associative learning to the study of prospective memory.

Judging the Importance of Constant and Variable Candidate Causes: A Test of the Power PC Theory

1998 ◽  
Vol 51 (1) ◽  
pp. 65-84 ◽  
Author(s):  
Frédéric Vallée-Tourangeau ◽  
Robin A. Murphy ◽  
Susan Drew ◽  
A.G. Baker
Keyword(s):  
Associative Learning ◽  
Causal Power ◽  
Base Rate ◽  
The Other ◽  
Associative Model ◽  
Contrast Model ◽  
Causal Induction ◽  
Power Theory ◽  
Factorial Combination ◽  
Target Effect

In two causal induction experiments subjects rated the importance of pairs of candidate causes in the production of a target effect; one candidate was present on every trial (constant cause), whereas the other was present on only some trials (variable cause). The design of both experiments consisted of a factorial combination of two values of the variable cause's covariation with the effect and three levels of the base rate of the effect. Judgements of the constant cause were inversely proportional to the level of covariation of the variable cause but were proportional to the base rate of the effect. The judgements were consistent with the predictions derived from the Rescorla-Wagner (1972) model of associative learning and with the predictions of the causal power theory of the probabilistic contrast model (Cheng, 1997) or “power PC theory”. However, judgements of the importance of the variable candidate cause were proportional to the base rate of the effect, a phenomenon that is in some cases anticipated by the power PC theory. An alternative associative model, Pearce's (1987) similarity-based generalization model, predicts the influence of the base rate of the effect on the estimates of both the constant and the variable cause.

scholarly journals Preprint Glautier (2017) Revised Non-local influences on associative learning: new data and further model evaluation

2017 ◽  
Author(s):  
Steven Glautier
Keyword(s):  
Maximum Likelihood ◽  
Model Selection ◽  
Associative Learning ◽  
Model Complexity ◽  
Simple Extension ◽  
Wagner Model ◽  
Akaike Weight ◽  
Learning Tasks ◽  
Non Local ◽  
Clear Support

Revised preprintPrevious work (Glautier, 2013) showed that the responses made by humans on trial n insimple associative learning tasks were influenced by events that took place on trial n−1and a simple extension of the Rescorla-Wagner Model (RWM Rescorla & Wagner,1972), the Memory Environment Cue Array (MECA) model, was presented to accountfor those results. In the current work further evidence of non-local influences onresponding during associative learning tasks is presented. The Rescorla-Wagner modeland the MECA model are evaluated as models for the observed data using qualitative,näive maximum likelihood, and Akaike weight analyses. In two experiments the Akaikeweight analyses strongly supported the simpler Rescorla-Wagner model over the MECAmodel but the qualitive and näive maximum likelihood analyses strongly supported theMECA model model over the simpler Rescorla-Wagner model. In Experiment 2 thisapparent conflict was resolved using a generalisation criterion test (Ahn, Busemeyer,Wagenmakers, & Stout, 2008; Busemeyer & Wang, 2000) which gave clear support tothe MECA model over the Rescorla-Wagner model. These results demonstrate thesuperiority of model selection using predictive validity, where possible, over selectionusing statistical adjustments for model complexity.

Taxing Attentional Resources Amplifies Cue Competition in Associative Learning

2012 ◽  
Author(s):  
Matthew J. Weber ◽  
Samuel B. Messing ◽  
Sharon L. Thompson-Schill
Keyword(s):  
Associative Learning ◽  
Cue Competition ◽  
Attentional Resources

Application of the New Method for the Rescorla-Wagner Model to a Probabilistic Learning Situation

2000 ◽  
Vol 87 (2) ◽  
pp. 413-414 ◽  
Author(s):  
Makoto Yamaguchi
Keyword(s):  
Base Rate ◽  
New Method ◽  
Experimental Paradigm ◽  
Learning Situation ◽  
Probabilistic Learning ◽  
Wagner Model ◽  
Base Rate Neglect ◽  
Alternative Method

The author recently showed that it is not always necessary to perform simulations to derive predictions from the Rescorla-Wagner model, and presented an alternative method. In this article, the method is applied to an influential experimental paradigm: base-rate neglect.

scholarly journals Preprint Glautier (2017) Non-local influences on associative learning: new data and further model evaluation

2017 ◽  
Author(s):  
Steven Glautier
Keyword(s):  
Maximum Likelihood ◽  
Model Selection ◽  
Associative Learning ◽  
Model Complexity ◽  
Simple Extension ◽  
Wagner Model ◽  
Akaike Weight ◽  
Learning Tasks ◽  
Non Local ◽  
Clear Support

There is a revised version at: https://osf.io/4nx2vPrevious work (Glautier, 2013) showed that the responses made by humans on trial n insimple associative learning tasks were influenced by events that took place on trial n−1and a simple extension of the Rescorla-Wagner Model (RWM Rescorla & Wagner,1972), the Memory Environment Cue Array (MECA) model, was presented to accountfor those results. In the current work further evidence of non-local influences onresponding during associative learning tasks is presented. The Rescorla-Wagner modeland the MECA model are evaluated as models for the observed data using qualitative,näive maximum likelihood, and Akaike weight analyses. In two experiments the Akaikeweight analyses strongly supported the simpler Rescorla-Wagner model over the MECAmodel but the qualitive and näive maximum likelihood analyses strongly supported theMECA model model over the simpler Rescorla-Wagner model. In Experiment 2 thisapparent conflict was resolved using a generalisation criterion test (Ahn, Busemeyer,Wagenmakers, & Stout, 2008; Busemeyer & Wang, 2000) which gave clear support tothe MECA model over the Rescorla-Wagner model. These results demonstrate thesuperiority of model selection using predictive validity, where possible, over selectionusing statistical adjustments for model complexity. Data and code available at Glautier(2017).

Sign in / Sign up

Export Citation Format

Share Document