Vencislav Popov

Building knowledge requires bricks, not sand: The critical role of familiar constituents in learning

Despite vast efforts to better understand human learning, some principles have been overlooked; specifically, that less familiar stimuli are more difficult to combine to create new knowledge and that this is because less familiar stimuli consume more working memory resources. Participants previously unfamiliar with Chinese characters were trained to discriminate visually similar characters during a visual search task over the course of a month, during which half of the characters appeared much more frequently. Ability to form associations involving these characters was tested via cued recall for novel associations consisting of two Chinese characters and an English word. Each week performance improved on the cued-recall task. Crucially, however, even though all Chinese character pairs were novel each week, those pairs consisting of more familiar characters were more easily learned. Performance on a working-memory task was better for more familiar stimuli, consistent with the claim that familiar stimuli consume fewer working memory resources. These findings have implications for optimal instruction, including second language learning.

The relational luring effect: Retrieval of relational information during associative recognition.

Here we argue that semantic relations (e.g., works in: nurse–hospital) have abstract independent representations in long-term memory (LTM) and that the same representation is accessed by all exemplars of a specific relation. We present evidence from 2 associative recognition experiments that uncovered a novel relational luring effect (RLE) in recognition memory. Participants studied word pairs, and then discriminated between intact (old) pairs and recombined lures. In the first experiment participants responded more slowly to lures that were relationally similar (table–cloth) to studied pairs (floor–carpet), in contrast to relationally dissimilar lures (pipe–water). Experiment 2 extended the RLE by showing a continuous effect of relational lure strength on recognition times (RTs), false alarms, and hits. It used a continuous pair recognition task, where each recombined lure or target could be preceded by 0, 1, 2, 3 or 4 different exemplars of the same relation. RTs and false alarms increased linearly with the number of different previously seen relationally similar pairs. Moreover, more typical exemplars of a given relation lead to a stronger RLE. Finally, hits for intact pairs also rose with the number of previously studied different relational instances. These results suggest that semantic relations exist as independent representations in LTM and that during associative recognition these representations can be a spurious source of familiarity. We discuss the implications of the RLE for current models of semantic and episodic memory, unitization in associative recognition, analogical reasoning and retrieval, as well as constructive memory research.

Fast Memory Integration Facilitated by Schema Consistency

Many of our everyday decisions are based not only on memories of direct experiences, but on memories that are integrated across multiple distinct experiences. Sometimes memory integration between existing memories and newly learnt information occurs rapidly, without requiring inference at the time of a decision. Such fast memory integration is known to be supported by the hippocampus but not the neocortex. In this study, we explore an alternative mechanism of fast memory integration, through related prior knowledge (i.e., schema), which is associated with neocortical learning. Paired associates were selected to be schema consistent or inconsistent, and confirmed with a latent semantic analysis of text corpora. We observed that after enabling fast learning by using material that is consistent with a schema, faster memory integration can occur. This result suggests that the hippocampus-mediated integration of new information is not the only available mechanism that supports fast memory integration.

Item strength affects working memory capacity

Do the processing and online manipulation of stimuli that are less familiar require more working memory (WM) resources? Is it more difficult to solve demanding problems when the symbols involved are less rather than more familiar? We explored these questions with a dual-task paradigm in which subjects had to solve algebra problems of different complexities while simultaneously holding novel symbol–digit associations in WM. The symbols were previously unknown Chinese characters, whose familiarity was manipulated by differential training frequency with a visual search task for nine hour-long sessions over 3 weeks. Subsequently, subjects solved equations that required one or two transformations. Before each trial, two different integers were assigned to two different Chinese characters of the same training frequency. Half of the time, those characters were present as variables in the equation and had to be substituted for the corresponding digits. After attempting to solve the equation, subjects had to recognize which two characters were shown immediately before that trial and to recall the integer associated with each. Solution accuracy and response times were better when the problems required one transformation only; variable substitution was not required; or the Chinese characters were high frequency. The effects of stimulus familiarity increased as the WM demands of the equation increased. Character–digit associations were also recalled less well with low-frequency characters. These results provide strong support that WM capacity depends not only on the number of chunks of information one is attempting to process but also on their strength or familiarity.

Inferential Pitfalls in Decoding Neural Representations.

A key challenge for cognitive neuroscience is to decipher the representational schemes of the brain. A recent class of decoding algorithms for fMRI data, stimulus-feature-based encoding models, is becoming increasingly popular for inferring the dimensions of neural representational spaces from stimulus-feature spaces. We argue that such inferences are not always valid, because decoding can occur even if the neural representational space and the stimulus-feature space use different representational schemes. This can happen when there is a systematic mapping between them, as shown by two simulations. In one simulation, we successfully decoded the binary representation of numbers from their decimal features. Since binary and decimal number systems use different representations, we cannot conclude that the binary representation encodes decimal features. In the second simulation, we successfully decoded the HSV color representation from the RGB representation of colors, even though these color spaces have different geometries and their dimensions have different interpretations. Detailed analysis of the predicted colors showed systematic deviations from the ground truth despite the high decoding accuracy, indicating that decoding accuracy on its own is not sufficient for making representational inferences. The same argument applies to the decoding of neural patterns from stimulus-feature spaces and we urge caution in inferring the nature of the neural code from such methods. We discuss ways to overcome these inferential limitations.

Cortical Networks Involved in Memory for Temporal Order

We examined the neurobiological basis of temporal resetting, an aspect of temporal order memory, using a version of the delayed-match-to-multiple-sample task. While in an fMRI scanner, participants evaluated whether an item was novel or whether it had appeared before or after a reset event that signified the start of a new block of trials. Participants responded “old” to items that were repeated within the current block and “new” to both novel items and items that had last appeared before the reset event (pseudonew items). Medial-temporal, prefrontal, and occipital regions responded to absolute novelty of the stimulus—they differentiated between novel items and previously seen items, but not between old and pseudonew items. Activation for pseudonew items in the frontopolar and parietal regions, in contrast, was intermediate between old and new items. The posterior cingulate cortex extending to precuneus was the only region that showed complete temporal resetting, and its activation reflected whether an item was new or old according to the task instructions regardless of its familiarity. There was also a significant Condition (old/pseudonew) × Familiarity (second/third presentations) interaction effect on behavioral and neural measures. For pseudonew items, greater familiarity decreased response accuracy, increased RTs, increased ACC activation, and increased functional connectivity between ACC and the left frontal pole. The reverse was observed for old items. On the basis of these results, we propose a theoretical framework in which temporal resetting relies on an episodic retrieval network that is modulated by cognitive control and conflict resolution.