The distinguishability between working memory (WM) and long-term memory has been a frequent and long-lasting source of debate in the literature. One recent method of identifying the relationship between the two systems has been to consider the influence of long-term memory effects, such as the levels-of-processing (LoP) effect, in WM. However, the few studies that have examined the LoP effect i...

Model compression is significant for the wide adoption of Recurrent Neural Networks (RNNs) in both user devices possessing limited resources and business clusters requiring quick responses to large-scale service requests. This work aims to learn structurally-sparse Long Short-Term Memory (LSTM) by reducing the sizes of basic structures within LSTM units, including input updates, gates, hidden s...

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient-based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, ...

Two experiments compared the effects of depth of processing on working memory (WM) and long-term memory (LTM) using a levels-of-processing (LOP) span task, a newly developed WM span procedure that involves processing to-be-remembered words based on their visual, phonological, or semantic characteristics. Depth of processing had minimal effect on WM tests, yet subsequent memory for the same item...

The change detection task has become a standard method for estimating the storage capacity of visual working memory. Most researchers assume that this task isolates the properties of an active short-term storage system that can be dissociated from long-term memory systems. However, long-term memory storage may influence performance on this task. In particular, memory traces from previous trials...

Performance on working memory (WM) tasks may partially be supported by long-term memory (LTM) processing. Hence, brain activation recently being implicated in WM may actually have been driven by (incidental) LTM formation. We examined which brain regions actually support successful WM processing, rather than being confounded by LTM processes, during the maintenance and probe phase of a WM task....

Available body of evidence concerning the relationship between insight problem solving and working memory (WM) is ambiguous. Several authors propose that restructuring of the problem representation requires controlled search processes, which needs planning and involvement of WM. Other researchers suggest that the restructuring is achieved through the automatic spread of activation in long-term ...

Evidence for long-term memory (LTM)-based control of attention has been found during the execution of highly practiced multi-step tasks. However, does LTM directly control for attention or are working memory (WM) processes involved? In the present study, this question was investigated with a dual-task paradigm. Participants executed either a highly practiced visuospatial sensorimotor task (spee...

The ability to associate stimuli across time and sensory modalities endows animals and humans with many of the complex, learned behaviors. For successful performance, associations need to be retrieved from long-term memory and maintained active in working memory. We investigated how this is accomplished in the avian brain. We trained carrion crows (Corvus corone) to perform a bimodal delayed pa...

The processing of numerical information induces a spatial response bias: Faster responses to small numbers with the left hand and faster responses to large numbers with the right hand. Most theories agree that long-term representations underlie this so called SNARC effect (Spatial Numerical Association of Response Codes; Dehaene et al., 1993). However, a spatial response bias was also observed ...