Gezinus Wolters

A Cortical Mechanism for Binding in Visual Working Memory

Luck and Vogel (1997) showed that the storage capacity of visual working memory is about four objects and that this capacity does not depend on the number of features making up the objects. Thus, visual working memory seems to process integrated objects rather than individual features, just as verbal working memory handles higher-order chunks instead of individual features or letters. In this article, we present a model based on synchronization and desynchronization of reverberatory neural assemblies, which can parsimoniously account for both the limited capacity of visual working memory, and for the temporary binding of multiple assemblies into a single pattern. A critical capacity of about three to four independent patterns showed up in our simulations, consistent with the results of Luck and Vogel. The same desynchronizing mechanism optimizing phase segregation between assemblies coding for separate features or multifeature objects poses a limit to the number of oscillatory reverberations. We show how retention of multiple features as visual chunks (feature conjunctions or objects) in terms of synchronized reverberatory assemblies may be achieved with and without long-term memory guidance.

Original Contribution: CALM: Categorizing and learning module

A new procedure (CALM: Categorizing and Learning Module) is introduced for unsupervised learning in modular neural networks. The work described addresses a number of problems in connectionist modeling, such as lack of speed, lack of stability, inability to learn either with or without supervision, and the inability to both discriminate between and generalize over patterns. CALM is a single module that can be used to construct larger networks. A CALM module consists of pairs of excitatory Representation- and inhibitory Veto-nodes, and an Arousal-node. Because of the fixed internal wiring pattern of a module, the Arousal-node is sensitive to the novelty of the input pattern. The activation of the Arousal-node determines two psychologically motivated types of learning operating in the module: elaboration learning, which implies a high learning rate and the distribution of nonspecific, random activations in the module, and activation learning, which has only base rate learning without random activations. The learning rule used is a modified version of a rule described by Grossberg. The workings of CALM networks are illustrated in a number of simulations. It is shown that a CALM module quickly reaches a categorization, even with new patterns. Though categorization and learning are relatively fast compared to other models, CALM modules do not suffer from excessive plasticity. They are also shown to be capable of both discriminating between and generalizing over patterns. When presented with a pattern set exceeding the number of Representation-nodes, similar patterns are assigned to the same node. Multi-modular simulations showed that with supervised learning an average of 1.6 presentations sufficed to learn the EXOR function. Moreover, an unsupervised learning version of the McClelland and Rumelhart model successfully simulated a word superiority effect. It is concluded that the incorporation of psychologically and biologically plausible structural and functional characteristics, like modularity, unsupervised (competitive) learning, and a novelty dependent learning rate, may contribute to solving some of the problems often encountered in connectionist modeling.

Blinks of the Mind: Memory Effects of Attentional Processes

If 2 words are presented successively within 500 ms, subjects often miss the 2nd word. This attentional blink reflects a limited capacity to attend to incoming information. Memory effects were studied for words that fell within an attentional blink. Unrelated words were presented in a modified rapid serial visual presentation task at varying stimulus-onset asynchronies, and attention was systematically manipulated. Subsequently, recognition, repetition priming, and semantic priming were measured separately in 3 experiments. Unidentified words showed no recognition and no repetition priming. However, blinked (i.e., unidentified) words did produce semantic priming in related words. When, for instance, ring was blinked, it was easier to subsequently identify wedding than apple. In contrast, when the blinked word itself was presented again, it was not easier to identify than an unrelated word. Possible interpretations of this paradoxical finding are discussed.

Quantification of Small Numbers of Dots: Subitizing or Pattern Recognition?

The fast and accurate quantification of small numbers of elements, called subitizing, has been explained by assuming either a limited capacity for immediate apprehension of numerosity, a fast number-dependent quantification process, or a pattern recognition process. A review of the available evidence suggests a pattern-recognition explanation for subitizing. To test this, patterns of 4-18 dots were presented in daily sessions on 5 consecutive days. There were two experimental conditions. In the constant-pattern condition, each number of dots was presented as the same random pattern throughout the experiment. In the variable-pattern condition, all patterns were generated randomly before each trial. By Day 5, both reaction times and accuracy scores in the constant-pattern condition indicated that subitizing occurred over the whole range of dot numbers used. Subitizing did not develop in the variable-pattern condition, although some practice effects were found. It is concluded that subitizing can best be explained as a patternrecognition process. Under normal circumstances this process is limited to small numbers of elements, because with larger numbers the number of possible configurations becomes too large to allow unequivocal pattern discrimination and pattern recognition.

Implicit memory for words presented during anaesthesia

Abstract Tests for implicit memory seem to be rather insensitive to the amount of attention given to stimuli during study. In the experiment reported here, the effect of a complete absence of attention during presentation of stimulus material on implicit memory performance was studied. Surgical patients were auditorily presented with exemplars of word categories during general anaesthesia. At the earliest convenient time after surgery, they were requested to generate category exemplars. Although none of the patients expressed any conscious recollection of events during the surgical episode, experimental patients generated significantly more critical exemplars than a control group. Apparently, information presented during anaesthesia can influence post-operative performance in an implicit memory task. Repetition priming seems possible even in the absence of attentional processing at the time of encoding. It is argued that the nature of the unattended encoding process can best be described as the automatic ...

The effects of advance location cueing on latencies in a single-letter recognition task

SummaryThis study investigates facilitating effects of foreknowledge of position in single-letter recognition tasks with latency as the dependent variable. Three kinds of cue are used: symbolic cues, location cues and neutral cues. For the effects of symbolic and location cues the following predictions are derived from the literature: increasing benefits with increasing SOAs for symbolic cues and constant benefits with increasing SOAs for location cues.Cue-detection times and letter-recognition times with the different cues were determined in three experiments. In the experiments the same general paradigm was used. In Experiment 1 cue-detection times appeared to differ as a function of cue eccentricity; in Experiments 2 and 3 cues were adjusted in such a way that cue-detection times were about the same. Cues reliably indicated the target position in Experiments 1 and 2; in Experiment 3 cue validity was reduced to 50%.The cueing results conformed partly to the predictions. All experiments suggested a small benefit under symbolic cueing and a larger benefit under location cueing. Except for the smallest SOA used, for both types of cue benefits were invariant over SOAs. The implications of these results in relation to other single-letter recognition studies are briefly discussed.

Full versus divided attention and implicit memory performance.

Effects of full and divided attention during study on explicit and implicit memory performance were investigated in two experiments. Study time was manipulated in a third experiment. Experiment 1 showed that both similar and dissociative effects can be found in the two kinds of memory test, depending on the difficulty of the concurrent tasks used in the divided-attention condition. In this experiment, however, standard implicit memory tests were used and contamination by explicit memory influences cannot be ruled out. Therefore, in Experiments 2 and 3 the process dissociation procedure was applied. Manipulations of attention during study and of study time clearly affected the controlled (explicit) memory component, but had no effect on the automatic (implicit) memory component. Theoretical implications of these findings are discussed.

Familiar face recognition as a function of distance and illumination: a practical tool for use in the courtroom

This paper describes a practical courtroom tool that provides insight into familiar face recognition accuracy as a function of distance and illumination. Subjects were shown pictures of faces of famous persons or lookalikes, with an exposure time of 12 seconds. Subjects were asked whether the person shown was familiar or unfamiliar to them. If familiarity was indicated, subjects had to identify the person. Seven distances (3–40 meters, i.e. 10–131 feet) and nine illumination levels (0.3–3000 lux) were combined, which resulted in a large matrix, each cell containing a hit score and a false alarm score. From these raw data diagnostic values were derived. The results show a systematic increase of recognition performance with decreasing distance and increasing illumination. The diagnostic values suggest a practical rule of thumb, the Rule of Fifteen: even in ideal conditions the desired diagnostic value of 15 is reached at no more than 15 m (49 ft), and no less than 15 lux.

Single-letter recognition accuracy benefits from advance cuing of location

With single-item visual displays, the facilitating effects of foreknowledge of item location have been demonstrated by using latency and accuracy as the dependent variables in detection tasks and by using latency as the dependent variable in recognition tasks. Evidence of such facilitat-ing effects obtained by using accuracy as the dependent variable in recognition tasks is scarce and rather ambiguous. One of the difficulties in demonstrating the beneficial effects of location cuing in recognition accuracy may be the fact that in these tasks a performance baseline of no cuing is used that leaves only relatively little room for improvement. Therefore, in the present experiments, we varied the performance baseline by presenting letters on one of three imaginary circles at different distances from the fixation point, and determined, for each subject, a presentation time that resulted in an overall no-cue accuracy rate of .75. In the experimental trials, three cue conditions were used: position cue (cuing 1 of 15 possible positions), ring cue (cuing 5 positions, all on one of the three circles), and no cue (cuing all 15 possible locations). In Experiment 1, the cue conditions were randomized. In Experiment 2, the cue conditions were blocked. Significant beneficial effects of foreknowledge of position were found in both experiments. Beneficial effects of ring cuing were found only in Experiment 2. In both experiments, the benefits of cuing are positively related to room for improvement. Partial correlations show that it is, indeed, room for improvement, and not the covarying factor of distance from fixation point, that determines cuing benefits. The theoretical implications of the beneficial effects of ring cuing under blocked presentation conditions are discussed.

Flashbulb and Event Memory of September 11, 2001: Consistency, Confidence and Age Effects:

We used the terrorist attacks on September 11, 2001 to investigate if the formation, confidence, and nature of flashbulb memories were dependent on age. In addition, we compared the consistency over time of flashbulb memories with event memory, i.e., widely publicized factual details of the event, in a group of young respondents. College students (n = 34, M age 24.8 yr.) were questioned 2 wk. after the attack and again 2 mo. later. At the later time, also a group of healthy elderly respondents (n = 20, M age 70.5 yr.) was asked the same questions. Performance of young and old participants did not differ. Flashbulb memories were found without exception for both time periods and in both age groups. These memories had high confidence ratings and were described as very vivid. The original event was judged to have been accompanied by high emotion and rehearsal. In the college group, event memory, and to a smaller extent also flashbulb memory, decreased in accuracy already over the 2 mo. We conclude that flashbulb memories are a special case of normal episodic memory for emotional events. The creation of flashbulb memories, however, requires a special scenario of emotional arousal and rehearsal.