Exploring interactions between long-term and working memory
using the Hebbian repetition effect
Kathryn L. Gigler & Paul J. Reber, Department of Psychology, Northwestern University, Evanston, IL
• WM is often referred to as a gateway for LTM,
though the systems also interact in the opposite
direction
Experimental results
Exp 1: Same
starting point for
repeated sequence
0.8
Accuracy (% Correct)
• This can be seen in the ability of experts to
maintain large quantities of domain-specific
information in mind, as LTM supports expanded
WM capacity for this information1
• The Hebb repetition effect can be seen as an in-lab
demonstration of this; participants perform better on
material that was presented repeatedly than on
material that was not in a WM task2
*
0.7
0.6
0.5
0.6
0.55
0.5
0.45
SeVi task and methods
Sequential visuospatial working memory task (SeVi-WM)4,5
Presentation Phase (A)
• Sequence of cues scroll vertically down to targets at bottom of screen
Response Phase (B)
• Participants reproduce the sequence using D/F/J/K keys
• Length of presented sequence is adaptive via staircase method
• 2 correct or incorrect leads to +1/-1 length for next sequence
12-item repeating sequence
DJFKJKDFJDKF
4-item fragment
DJFKJKDFJDKF
5-item fragment
DJFKJKDFJDKF
6-item fragment
DJFKJKDFJDKF
B
4-item fragment
DJFKJKDFJDKF
5-item fragment
DJFKJKDFJDKF
6-item fragment
DJFKJKDFJDKF
7
6
5
4
3
2
1
0
Unaware
Hebb repetition effect:
• In Exp 1, where the repeated sequence was
obvious to participants, the effect can be
observed at adaptive sequence lengths in all
participants, F(39)=10.61, p<.01
• In Exp 2 and 3, where the repeated sequence
was not obvious to participants, the effect is
not observed across all participants
A
8
0.4
0.3
• What is the relationship between LTM,
awareness, and WM under these conditions?
Experiments 2 and 3 (B):
• Presented fragments start from randomized
points within the repeated sequence
• Exp 2: Two 50-minute training sessions
20% repeated and 80% random (N=26)
• Exp 3: Two 50-minute training sessions
40% repeated and 60% random (N=21)
*
0.4
• Whether this effect can be observed when the
repeated material is presented in fragments, and
whether the effect occurs with or without explicit
awareness, remain controversial3
• Can the Hebb repetition effect be observed in a
novel task where the repeated sequence is
presented in fragments?
Experiment 1 (A):
• Presented fragments always start from the
same point within the repeated sequence
• Two 50-minute training sessions; 20%
repeated/80% random (N=40)
Repetition effect across recognition groups
Exp 2 and 3: Randomized
starting point for repeated
sequence
WM span on task (# of items)
Introduction
However, in Exp 3, the repetition effect is observed only in participants able to recognize the repeated sequence during a post-training
recognition task, F(19)=4.86, p <.05
Accuracy (% Correct)
•
Aware
Effect of LTM/awareness:
• In Exp 2, no participants were aware of the
repeated sequence, as measured on a
post-training test of recognition
• In Exp 3, a subset of participants identified
the repeated sequence, and these
participants only demonstrated the
repetition effect, F(19)=4.86, p <.05
Unaware
Aware
• Additionally, higher initial
WM span seems to lead to
greater LTM awareness, as
higher-span participants
recognized the repeated
sequence at post-training, t
(10)=2.73, p <.03
Conclusions
• The repetition effect was observed (in some
participants) at a presentation rate of 40%
repeated/80% random, but not at 20%
repeated/80% random
• Suggests that sequences must be of a
certain length in order to learn statistical
regularities between items when sequence is
fragmented
• As measured by the repetition effect and
recognition test, advantage for repeated material
occurred only if participants were aware of the
repeated sequence
• This advantage was enhanced for higherspan participants, indicating a “rich get
richer” effect
Higher WM span
Hebb repetition effect
(supported by LTM)
Enhanced WM on task
Enhanced task
performance
References
1. Ericsson, K.A. & Kintsch, W. (1995). Long-term working memory. Psychol Rev, 102(2):211-45.
2. Hebb, D.O. (1961). Distinctive features of learning in the higher animal. In J.F. Delafresnaye (Ed.), Brain mechanisms and learning, 37-46. Oxford:
Blackwell.
3. Couture, M. & Tremblay, S. (2006). Exploring the characterisitcs of the visuospatial Hebb repetition effect. Memory & Cognition, 34(8), 1720-9.
4. Sanchez, D.J., Gobel, E.W. & Reber, P.J. (2010). Performing the unexplainable: Implicit task performance reveals individually reliable sequence learning
without explicit knowledge. Psych Bull & Review, 17(6), 790-6.
5. Gigler, K.L. & Reber, P.J. Sequence-specific and non-specific gains in working memory. Under revision.
Acknowledgements: Funding for this work was provided by the Department of Psychology and the Family Institute of Northwestern University.