Spanning seven orders of magnitude: a challenge for

The prefrontal cortex has long been thought to subserve both working memory (the holding of information online for processing) and executive functions (deciding how to manipulate working memory and perform processing). Although many computational models of working memory have been developed, the mechanistic basis of executive function remains elusive, often amounting to a homunculus. This article presents an attempt to deconstruct this homunculus through powerful learning mechanisms that allow a computational model of the prefrontal cortex to control both itself and other brain areas in a strategic, task-appropriate manner. These learning mechanisms are based on subcortical structures in the midbrain, basal ganglia, and amygdala, which together form an actor-critic architecture. The critic system learns which prefrontal representations are task relevant and trains the actor, which in turn provides a dynamic gating mechanism for controlling working memory updating. Computationally, the learning mechanism is designed to simultaneously solve the temporal and structural credit assignment problems. The model’s performance compares favorably with standard backpropagation-based temporal learning mechanisms on the challenging 1-2-AX working memory task and other benchmark working memory tasks.

Computational models have emerged as a powerful tool for studying the complex task of driving, allowing researchers to simulate driver behavior and explore the parameters and constraints of this behavior. In this paper we investigate the advantages of developing rigorous computational models of driver behavior in cognitive architectures — computational frameworks with underlying psychological theories that incorporate basic properties and limitations of the human system. In particular, we describe an integrated driver model developed in the ACT-R cognitive architecture and demonstrate how this model accounts for the steering profiles, lateral-position profiles, and gaze distributions of human drivers during lane keeping, curve negotiation, and lane changing. The model has implications both for theoretical accounts of complex dynamic tasks in the context of cognitive architectures and for practical applications in predicting and recognizing driver behavior and distraction.

The fan effect (Anderson, 1974) has been attributed to interference among competing associations to a concept. Recently, it has been suggested that such effects might be due to multiple mental models (Radvansky, Spieler, & Zacks, 1993) or suppression of concepts (Anderson & Spellman, 1995; Conway & Engle, 1994). We show that the ACT-R (Adaptive Control of Thought-Rational) theory, which embodies associative interference, is consistent with the Radvansky et al results and we fail to find any evidence for concept suppression in a new fan experiment. The ACT-R model provides good quantitative fits to the results from a variety of experiments. The three key concepts in these fits are (a) the associative strength between two concepts reflect the degree to which one concept predicts the other; (b) foils are rejected by retrieving mismatching facts; and (c) subjects can adjust the relative weights they give to various cues in retrieval.

Many theories of skill acquisition have had considerable success in addressing the fine details of learning in relatively simple tasks, but can they scale up to complex tasks that are more typical of human learning in the real world? Some theories argue for scalability by making the implicit assumption that complex tasks consist of many smaller parts, which are learned according to basic learning principles. Surprisingly, there has been rather sparse empirical testing of this crucial assumption. In this article, we examine this assumption directly by decomposing the learning in the Kanfer–Ackerman Air-Traffic Controller Task (Ackerman, 1988) from the learning at the global level all the way down to the learning at the keystroke level. First, we reanalyze the data from Ackerman (1988) and show that the learning in this complex task does indeed reflect the learning of smaller parts at the keystroke level. Second, in a follow-up eye-tracking experiment, we show that a large portion of the learning at the keystroke level reflects the learning even at a lower, i.e., attentional level. © 2001 Academic Press Over the past 2 decades there have appeared a number of theories of skill

predictions of exemplar models and that supported prototype models. In the authors ’ view, this evidence confounded the issue of the nature of the category representation with the type of response rule (probabilistic vs. deterministic) that was used. Also, their designs did not test whether the prototype models correctly predicted generalization performance. The present work demonstrates that an exemplar model that includes a response-scaling mechanism provides a natural account of all of Smith et al.’s experimental results. Furthermore, the exemplar model predicts classification performance better than the prototype models when novel transfer stimuli are included in the experimental designs. A classic issue in cognitive psychology concerns the manner in which people represent categories in memory. According to prototype models (Homa, 1984; Posner & Keele, 1968; Reed, 1972), people represent categories by forming a summary representation that is a central tendency of all of the experienced members of a

An experiment investigates (1) how the physical structure of a computer screen layout affects visual search and (2) how people select a found target object with a mouse. Two structures are examinedlabeled visual hierarchies (groups of objects with one label per group) and unlabeled visual hierarchies (groups without labels). Search and selection times were separated by imposing a point-completion deadline that discouraged participants from moving the mouse until they found the target. The observed search times indicate that labeled visual hierarchies can be searched much more efficiently than unlabeled visual hierarchies, and suggest that people use a fundamentally different strategy for each of the two structures. The results have implications for screen layout design and cognitive modeling of visual search. The observed mouse pointing times suggest that people use a slower and more accurate speed-accuracy operating characteristic to select a target with a mouse when visual distractor...

Participants memorized briefly presented sets of digits, a subset of which had to be accessed as input for arithmetic tasks (the active set), whereas another subset had to be remembered independently of the concurrent task (the passive set). Latencies for arithmetic operations were a function of the setsize of active but not passive sets. Object-switch costs were observed when successive operations were applied to different digits within an active set. Participants took2stoencode a passive set so that it did not affect processing latencies (Experiment 2). The results support a model distinguishing 3 states of representations in working memory: the activated part of long-term memory, a capacity limited region of direct access, and a focus of attention. Working memory is commonly described as a system for simultaneous storage and processing of information. The relation between “storage ” and “processing, ” however, is rarely specified. Resource models generally posit a common resource (e.g., activation) that must be shared between the two functions (Just & Carpenter, 1992). Evidence from dual task studies, however, casts doubt on the resource-sharing hypothesis: There are numerous examples in the literature of processing that is largely unimpaired by a concurrent short-term memory demand, even when the memory demand is close to the maximum span (e.g., Foos & Wright,

Unexpected interruptions introduced during the execution phase of simple Tower of London problems incurred a time cost when the interrupted goal was retrieved, and this cost was exacerbated the longer the goal was suspended. Furthermore, time taken to retrieve goals was greater following a more complex interruption, indicating that processing limitations may be as important as time-based limitations in determining the ease of goal retrieval. Such findings cannot simply be attributed to task-switching costs and are evaluated in relation to current models of goal memory (E. M. Altmann & G. J. Trafton, 2002; J. R. Anderson & S. Douglass, 2001), which provide a useful basis for the investigation and interpretation of interruption effects.

Measures of entropy are useful for explaining the behaviour of cognitive models. We demonstrate that entropy can not only help to analyse the performance of the model, but also it can be used to control model parameters and improve the match between the model and data. We present a cognitive model that uses local computations of entropy to moderate its own behaviour and matches the data fairly well.