Skip to main content
SpringerLink
Log in

A Neural Network Model of Episode Representations in Working Memory

  • Published:
Cognitive Computation Aims and scope Submit manuscript

Abstract

We present a neural network model of the storage of episode representations in working memory (WM). Our key idea is that episodes are encoded in WM as prepared sensorimotor routines, i.e. as prepared sequences of attentional and motor operations. Our network reproduces several experimental findings about the representation of prepared sequences in prefrontal cortex. Interpreted as a model of WM episode representations, it has useful applications in an account of long-term memory for episodes and in accounts of sentence processing.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Similar content being viewed by others

Notes

  1. Manuscript in preparation, Knott and Takac: Locomotion actions as sequentially structured sensorimotor routines.

  2. In fact, even when the network is used for perception of episodes, the selection mechanism has an impact on the agent’s behaviour. The expected episode is a sequence whose first two items are planned attentional actions: during perception, these actions will actually be executed by the agent, with results that depend on the world as well as on the agent’s expectations and which might well result in revisions to the selected episode. The model thus allows for active perceptual operations during the process of selecting an episode, creating the structural coupling between its representational system and the environment that is characteristic of embodied systems (see, for example, [32]).

  3. The most obvious kind of ‘attentional actions’ are overt movements, such as saccades. But attentional actions also involve cognitive operations, in particular top−down activation of semantic representations. These top−down activation operations can encode either the expected result of a forthcoming object classification process [34] or the category of properties of a desired search target [35, 36]. Our attentional actions represent objects in the sense that they represent expected or sought-for object categories.

  4. Since WM representations are normally understood to be ‘maintained in the face of incoming perceptual stimuli’, we have to assume a special operation to remove the current episode representation from the dynamic episodic buffer before the next episode begins—an operation that probably involves an element of self-inhibition (see, for example, Mayr and Keele [39]). In previous work [19], we have considered the nature of this operation, but in the present study, we just use the end-of-episode signal to stand in for this operation.

  5. If the predicted episode ranked among multiple episodes predicted by the VLMM with equal frequency, e.g. occupying 2nd–4th position, its rank would be the upper bound, i.e. 2.

  6. Candidates were determined by top−down reconstruction, i.e. replayed as a temporal sequence in the aggregate SM signal layer.

  7. The distributed model performed better by 3.7 % in total grammaticality, 0.2 % in compatibility, 0.8 % in matches, 0.09 in rank, and 5.1 % in rank base [37].

References

  1. Baddeley A, Hitch G. Working memory. In: Bower G, editor. The psychology of learning and motivation: advances in research and theory (Vol. 8). New York: Academic Press; 1974. p. 47–90.

  2. Baddeley A. The episodic buffer: a new component of working memory? TICS. 2000;4(11):417–23.

    Google Scholar 

  3. Tulving E, Markowitsch H. Episodic and declarative memory: role of the hippocampus. Hippocampus. 1998;8:198–204.

    Article  CAS  PubMed  Google Scholar 

  4. Abraham W, Logan B, Greenwood J, Dragunow M. Induction and experience-dependent consolidation of stable long-term potentiation lasting months in the hippocampus. J Neurosci. 2002;22:9626–34.

    CAS  PubMed  Google Scholar 

  5. Levelt W. Speaking: from intention to articulation. Cambridge: MIT Press; 1989.

    Google Scholar 

  6. Haarmann H, Cameron K, Ruchkin D. Short-term semantic retention during on-line sentence comprehension. Brain potential evidence from filler-gap constructions. Cognit Brain Res. 2003;15:178–90.

    Article  Google Scholar 

  7. Mayberry M, Miikkulainen R. Incremental nonmonotonic sentence interpretation through semantic self-organization. The University of Texas at Austin: Department of Computer Sciences; 2008; AI08–12.

  8. Ballard D, Hayhoe M, Pook P, Rao R. Deictic codes for the embodiment of cognition. Behav Brain Sci. 1997;20(4):723–67.

    CAS  PubMed  Google Scholar 

  9. Knott A. Sensorimotor cognition and natural language syntax. Cambridge: MIT Press; 2012.

    Google Scholar 

  10. Knott A. Syntactic structures as descriptions of sensorimotor processes. Biolinguistics. 2014;8:1–52.

    Google Scholar 

  11. Wallenstein G, Eichenbaum H, Hasselmo M. The hippocampus as an associator of discontiguous events. Trends Neurosci. 1998;21:317–23.

    Article  CAS  PubMed  Google Scholar 

  12. Lee A, Wilson M. Memory of sequential experience in the hippocampus during slow wave sleep. Neuron. 2002;36:1183–94.

    Article  CAS  PubMed  Google Scholar 

  13. Cutsuridis V, Hasselmo M. Spatial memory sequence encoding and replay during modeled theta and ripple oscillations. Cognit Comput. 2011;3:554–74.

    Article  Google Scholar 

  14. Haggard P. Human volition: towards a neuroscience of will. Nat Rev Neurosci. 2008;9:934–46.

    Article  CAS  PubMed  Google Scholar 

  15. Tessitore G, Prevete R, Catanzariti E, Tamburrini G. From motor to sensory processing in mirror neuron computational modelling. Biol Cybern. 2010;103:471–85.

    Article  PubMed  Google Scholar 

  16. Chambon V, Wenke D, Fleming S, Prinz W, Haggard P. An online neural substrate for sense of agency. Cereb Cortex. 2012;23:1031–7.

    Article  PubMed  Google Scholar 

  17. Tipper S, Lortie C, Baylis G. Selective reaching: evidence for action-centered attention. J Exp Psychol Hum Percept Perform. 1992;18:891–905.

    Article  CAS  PubMed  Google Scholar 

  18. Johansson R, Westling G, Backstrom A, Flanagan J. Eye-hand coordination in object manipulation. J Neurosci. 2001;21(17):6917–32.

    CAS  PubMed  Google Scholar 

  19. Caza G, Knott A. Pragmatic bootstrapping: a neural network model of vocabulary acquisition. Lang Learn Dev. 2012;8:1–23.

    Article  Google Scholar 

  20. Knott A. How infants learn word meanings and propositional attitudes: a neural network model. In: Hung TW, editor. Lang action. Berlin: Springer; 2014. p. 107–24.

    Google Scholar 

  21. Lee-Hand J, Knott A. A model of causative actions: from motor learning to syntactic structure. In: Proceedings of the 35th annual meeting of the cognitive science society. Berlin; 2013; p. 2849–2854.

  22. Curtis C, D’Esposito M. Persistent activity in the prefrontal cortex during working memory. Trends Cognit Sci. 2003;7(9):415–23.

    Article  Google Scholar 

  23. Barsalou L. Grounded cognition. Ann Rev Psychol. 2008;59:617–45.

    Article  Google Scholar 

  24. Glenberg A, Gallese V. Action-based language: a theory of language acquisition, comprehension, and production. Cortex. 2012;48:905–22.

    Article  PubMed  Google Scholar 

  25. Tulving E. Elements of episodic memory. New York: Oxford University Press; 1983.

    Google Scholar 

  26. Barone P, Joseph JP. Prefrontal cortex and spatial sequencing in macaque monkey. Exp Brain Res. 1989;78:447–64.

    Article  CAS  PubMed  Google Scholar 

  27. Averbeck B, Chafee M, Crowe D, Georgopoulos A. Parallel processing of serial movements in prefrontal cortex. PNAS. 2002;99(20):13172–7.

    Article  PubMed Central  CAS  PubMed  Google Scholar 

  28. Rhodes B, Bullock D, Verwey W, Averbeck B, Page M. Learning and production of movement sequences: behavioral, neurophysiological, and modeling perspectives. Hum Mov Sci. 2004;23:699–746.

    Article  PubMed  Google Scholar 

  29. Averbeck B, Sohn J, Lee D. Activity in prefrontal cortex during dynamic selection of action sequences. Nat Neurosci. 2006;9(2):276–82.

    Article  CAS  PubMed  Google Scholar 

  30. Shallice T, Burgess P. Deficits in strategy application following frontal lobe damage in man. Brain. 1991;114:727–41.

    Article  PubMed  Google Scholar 

  31. Kaller C, Heinze K, Frenkel A, Läppchen C, Unterrainer J, Weiller C, et al. Differential impact of continuous theta-burst stimulation over left and right DLPFC on planning. Hum Brain Mapp. 2013;34:36–51.

    Article  PubMed  Google Scholar 

  32. Ziemke T. What’s that thing called embodiment? In: Alterman R, Kirsh D, editors. Proceedings of the 25th Annual Conference of the Cognitive Science Society. Mahwah, NJ: Lawrence Erlbaum; 2003. p. 1134–9.

  33. Kohonen T. Self-organized formation of topologically correct feature maps. Biol Cybern. 1982;43:59–69.

    Article  Google Scholar 

  34. Kahneman D, Treisman A, Gibbs B. The reviewing of object files: object-specific integration of information. Cognit Psychol. 1992;24:175–219.

    Article  CAS  PubMed  Google Scholar 

  35. Hasegawa R, Matsumoto M, Mikami A. Search target selection in monkey prefrontal cortex. J Neurophysiol. 2000;84:1692–6.

    CAS  PubMed  Google Scholar 

  36. Fix J, Rougier N, Alexandre F. A dynamic neural field approach to the covert and overt deployment of spatial attention. Cognit Comput. 2011;3:279–93.

    Article  Google Scholar 

  37. Takac M, Knott A. A revised neural network model of episode representations in working memory. Dept of Computer Science, University of Otago; 2014; OUCS-2014-03.

  38. Strickert M, Hammer B. Merge SOM for temporal data. Neurocomputing. 2005;64:39–71.

    Article  Google Scholar 

  39. Mayr U, Keele S. Changing internal constraints on action: the role of backward inhibition. J Exp Psychol Gen. 2000;129(1):4–26.

    Article  CAS  PubMed  Google Scholar 

  40. Vickers D, Lee MD. Dynamic models of simple judgments: II. Properties of a self-organizing PAGAN (parallel, adaptive, generalized accumulator network) model for multi-choice tasks. Nonlinear Dyn Psychol Life Scie. 2000;4(1):1–31.

    Article  Google Scholar 

  41. James DL, Miikkulainen R. SARDNET: a self-organizing feature map for sequences. In: Tesauro G, Touretzky DS, Leen TK, editors. Advances in neural information processing systems 7. Cambridge: MIT Press; 1995. p. 577–84.

    Google Scholar 

  42. Miikkulainen R. Subsymbolic case-role analysis of sentences with embedded clauses. Cognit Sci. 1996;20:47–73.

    Article  Google Scholar 

  43. Chang F. Symbolically speaking: a connectionist model of sentence production. Cognit Sci. 2002;26:609–51.

    Article  Google Scholar 

  44. Rohde D. A connectionist model of sentence comprehension and production. School of Computer Science. Carnegie Mellon University; 2002.

  45. Takac M, Benuskova L, Knott A. Mapping sensorimotor sequences to word sequences: a connectionist model of language acquisition and sentence generation. Cognition. 2012;125:288–308.

    Article  PubMed  Google Scholar 

  46. Collins M. A new statistical parser based on bigram lexical dependencies. In: Proceedings of the 34th meeting of the ACL. Santa Cruz; 1996; p. 184–191.

  47. Dispaldro M, Leonard L, Corradi N, Ruffino M, Bronte T, Facoetti A. Visual attentional engagement deficits in children with specific language impairment and their role in real-time language processing. Cortex. 2013;49:2126–39.

    Article  PubMed Central  PubMed  Google Scholar 

  48. Balconi M. Dorsolateral prefrontal cortex, working memory and episodic memory processes: insight through transcranial magnetic stimulation techniques. Neurosci Bull. 2013;29(3):381–9.

    Article  PubMed  Google Scholar 

  49. Burgess N, Hitch G. Memory for serial order: a network model of the phonological loop and its timing. Psychol Rev. 1999;106:551–81.

    Article  Google Scholar 

  50. Burgess N, Hitch G. A revised model of short-term memory and long-term learning of verbal sequences. J Mem Lang. 2006;55(4):627–52.

    Article  Google Scholar 

  51. Brown GDA, Hulme C, Preece T. Oscillator-based memory for serial order. Psychol Rev. 2000;107:127–81.

    Article  CAS  PubMed  Google Scholar 

  52. Howard M, Kahana M. A distributed representation of temporal context. J Math Psychol. 2002;46:269–99.

    Article  Google Scholar 

  53. Page M, Norris D. A model linking immediate serial recall, the Hebb repetition effect and the learning of phonological word forms. Philos Trans R Soc B. 2009;364:3737–53.

    Article  CAS  Google Scholar 

  54. Botvinick MM, Plaut DC. Short-term memory for serial order: a recurrent neural network model. Psychol Rev. 2006;113(2):201–33.

    Article  PubMed  Google Scholar 

  55. Dominey P, Arbib M, Joseph JP. A model of corticostriatal plasticity for learning associations and sequences. J Cognit Neurosci. 1995;7(3):311–36.

    Article  CAS  Google Scholar 

  56. Dominey P, Hoen M, Inui T. A neurolinguistic model of grammatical construction processing. J Cognit Neurosci. 2006;18(12):2088–107.

    Article  Google Scholar 

  57. Shastri L. Episodic memory trace formation in the hippocampal system: a model of cortico-hippocampal interaction. Berkeley: International Computer Science Institute (ICSI), UC Berkeley; 2001; TR-01-004.

  58. van der Velde F, de Kamps M. Neural blackboard architectures of combinatorial structures in cognition. Behav Brain Sci. 2006;29:37–108.

    PubMed  Google Scholar 

  59. Stewart T, Eliasmith C. Compositionality and biologically plausible models. In: Werning M, Hinzen W, editors. The Oxford handbook of compositionality. New York: Oxford University Press; 2012.

    Google Scholar 

  60. Dennis S. A memory-based theory of verbal cognition. Cognit Sci. 2005;29(2):145–93.

    Article  Google Scholar 

  61. Reynolds J, Zacks J, Braver T. A computational model of event segmentation from perceptual prediction. Cognit Sci. 2007;31:613–43.

    Article  Google Scholar 

Download references

Acknowledgments

This research was supported by NZ Marsden Fund, and partially supported by grants VEGA 1/0898/14 and KEGA 076UK-4/2013 for Martin Takac. We are grateful to Lubica Benuskova and Igor Farkas for helpful discussions.

Author information

Authors and Affiliations

  1. Department of Computer Science, University of Otago, PO Box 56, 9054, Dunedin, New Zealand

    Martin Takac & Alistair Knott

  2. Centre for Cognitive Science FMFI UK, Comenius University, Mlynska Dolina, 842 48, Bratislava, Slovakia

    Martin Takac

Authors
  1. Martin Takac
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Alistair Knott
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Martin Takac.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Takac, M., Knott, A. A Neural Network Model of Episode Representations in Working Memory. Cogn Comput 7, 509–525 (2015). https://doi.org/10.1007/s12559-015-9330-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s12559-015-9330-3

Keywords

Search

Navigation