Center for the Study of Brain, Mind and Behavior (CSBMB)

PDF version

Retrieval-induced forgetting (RIF) refers to the finding that retrieving a memory can impair subsequent recall of related memories. Here, we present a new model of how the brain gives rise to RIF in both semantic and episodic memory. The core of the model is a recently developed neural network learning algorithm that leverages regular oscillations in feedback inhibition to strengthen weak parts of target memories and to weaken competing memories. We use the model to address several puzzling findings relating to RIF including: why retrieval practice leads to more forgetting than simply presenting the target item; how RIF is affected by the strength of competing memories and the strength of the target (to-be-retrieved) memory; and why RIF sometimes generalizes to "independent cues", and sometimes does not. For all of these questions, we show that the model can account for existing results, and we generate novel predictions regarding boundary conditions on these results.

PDF version

We consider the effects of signal ‘sharpness’ or acuity on the performance of neural models of decision making. In these models, a vector of signals is presented and the subject must decide which of the elements of the vector is the largest. In an earlier paper (McMillen and Holmes,2006) we derived asymptotically optimal tests under the assumption that the elements of the signal vector were all equal except one, which was a quantity a greater than the rest. In this paper we consider the case of signals spread around a peak. The acuity is a measure of how strongly peaked the signal is. We find that the best test is one in which the detectors are multiplied by a matrix containing the possible signal vectors, and comparisons made on the resulting product. This matrix multiplication is a way to encode knowledge of the shape of the incoming signals. The advantage to this method, in terms of reducing the time to make a decision under error constraints, is correlated to the spread of the signals. The greater the spread of the signals around the peak, the greater is the advantage to performing the matrix multiplication, which can result in orders of magnitude improvement in decision times.

PDF version

Internalizing Representations of Correct Behavior From Feedback In a series of computational simulations we have demonstrated that a mechanism for error detection in reinforcement learning tasks (Holroyd & Coles, 2002) can also detect errors in speeded response tasks (Holroyd & Coles, 2002; Holroyd, Yeung, Coles & Cohen, 2005). The architecture of this system is based on the “actor-critic” architecture shown in figure 1, in which the behavior of a task module system (the “actor”) is evaluated by a separate monitoring system(the “critic”; Sutton & Barto, 1998). Our original simulation demonstrated that this architecture could use feedback to learn internal representations of appropriate behavior, such that the model could learn to perform tasks in the absence of external feedback (Holroyd & Coles, 2002). This work also demonstrated that the amplitudes of the feedback error-related negativity (ERN) and the response ERN, both components of the event-related brain potential (ERP) associated with error commission, respectively decreased and increased as subjects gradually came to rely on internal rather than external sources of information to evaluate their performance (for review of the ERN see Holroyd, Nieuwenhuis, Mars & Coles, 2004).

PDF version

"*Technical report 06-01 has been superseded by technical report 07-01 (above)."

PDF version

We present a new learning algorithm that leverages oscillations in the strength of neural inhibition to train neural networks. Raising inhibition can be used to identify weak parts of target memories, which are then strengthened. Conversely, lowering inhibition can be used to identify competitors, which are then punished. To update weights, we use the Contrastive Hebbian Learning equation, applied to successive time steps of the network. The sign of the weight change equation varies as a function of the phase of the inhibitory oscillation. We use the learning algorithm to account for behavioral data regarding how competition at retrieval affects subsequent memory. We also show that the learning algorithm's capacity for storing patterns increases steadily as a function of network size, and that the learning algorithm can memorize large numbers of correlated patterns without collapsing. Finally, we discuss how this work relates to neural data on theta oscillations and learning.

PDF version

We show that adaptive gain changes, by hypothesis mediated by the locus coeruleus (LC), can help optimize performance on simulated sensory discrimination tasks, even when no knowledge of stimulus timing is assumed. The metric of performance used here is the rate of correct responses (or 'reward rate') achieved by the simulated decision network. The primary model that we study has two layers: the first integrates sensory input directly and the second accumulates this filtered input (as well as noise from other brain areas) and translates it into motor responses via threshold crossing events. Gain transients occur after a physiologically-motivated delay following threshold crossings in the first layer - in this sense, gain schedules adapt according to accumulated sensory information. We adopt a linearization and reduction of the two layer model that allows a clear understanding of parameter effects and which is used to obtain a simpler set of optimization problems without sacrificing the generality of their solutions. By comparing optimal model reward rates in the presence of simulated LC-mediated gain changes with the (separately) optimized reward rates in the absence of such gain changes, the extent to which these gain changes contribute to enhanced task performance is determined, all for a 'standard parameter set' derived from fits to experiments. The results indicate that a significant improvement in reward (12-24%) is attributable to the LC-mediated gain mechanism. Additionally, the statistical variations in the optimal model gain transients from trial-to-trial agree with trends reported in recent experimental studies involving direct recordings from the LC. This provides converging evidence for the hypothesis that the LC plays a part in optimizing the dynamics of simple decision tasks.

PDF version

In this report we collect and summarize mathematical results and formalism appropriate to describing one-dimensional drift-diffusion processes (stochastic ordinary differential equations) and related first passage and probability density evolution problems, governed by the backward and forward Kolmogorov (Fokker-Planck) equations respectively. We start by reviewing the Neyman-Pearson and Sequential Probability Ratio tests as optimal strategies for choosing between two alternative hypotheses in the presence of accumulating, noisy data. The continuum analog of both of these tests is a constant drift-diffusion process, and we give direct proofs of optimality with respect to reward rate of such a process in the broader class of Ornstein-Uhlenbeck processes, both in terms of first passages and density evolution. These correspond to the free response and interrogation protocols used in psychological testing. We end by considering the effects of variable gain on selected inputs to drift-diffusion and Ornstein-Uhlenbeck processes, and deriving optimal gain schedules for time-varying signal-to-noise ratios.

PDF version

We have recently presented empirical evidence for the view that the NoGo-N2, an event-related brain potential (ERP) component observed in Go/NoGo tasks, reflects conflict arising from competition between the execution and inhibition of a single response (Nieuwenhuis, Yeung, Van den Wildenberg, & Ridderinkhof, in press). Furthermore, the source of the NoGo-N2 was localized to anterior cingulate cortex (ACC), a brain region thought to be involved in response conflict detection. Here we show that a recently proposed model of behavioral performance and ACC functioning in the Go/NoGo task can account for many aspects of the empirical data reported in Nieuwenhuis et al. We also discuss the aspects of the data that the model could not capture.

PDF version

This report presents a method for finding parameters of connectionist models that allows the behavior of the model to be fit as closely as possible to empirical data regarding the behavior of human subjects in psychological experiments. The method is based on minimization of a cost function that expresses how different the statistics describing behavior of the model are from the statistics of the subjects' performance in the experiment. An optimization algorithm is used to find the values of the parameters for which the value of the cost function is the smallest. The cost function also indicates whether the model's statistics are significantly different from those obtained in the experiment. In some cases, the method can find the required parameters automatically. In other cases it may help and accelerate the process of manual parameterization. The method has been implemented in Matlab and is fully documented. Code and examples are available for free download.