Parallel or serial, fast or slow, Testing theories of WMC online or offline? Testing theories of WMC Mark Nieuwenstein
In this talk: Phenomenon: Extended sparing in RSVP 2. Theories of extended sparing in RSVP 3. A qualitative test of the theories 4. A more general point Conclusions: WMC can be interrupted by a switch in central processing requirements, but not by a mask or another to-be-encoded item WMC occurs in parallel for successive items in RSVP WMC takes about 400 ms for a single letter in RSVP Occurrence of interruption of WMC depends on the likelihood of interference during WMC.
Phenomenon: Extended sparing in RSVP Nieuwenstein & Potter (Psychological Science, 2006)
Phenomenon: Extended sparing in RSVP Nieuwenstein & Potter (Psychological Science, 2006)
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Attentional Blink: T1 [blank ISI] T2 Mask Attention * T1 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Extended sparing T1 T2 T3 T4 Attention * T1 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Extended sparing T1 T2 T3 T4 Attention * T2 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Extended sparing T1 T2 T3 T4 Attention * T3 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Extended sparing T1 T2 T3 T4 Attention * T4 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Attention * T4 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011 2. Theories of extended sparing in RSVP: eSTST WMC Attention * T4 Wyble, Potter, Bowman, & Nieuwenstein, JEP:HPP 2009, JEP:Gen 2011
2. Other theories of extended sparing in RSVP Serial Consolidation (Taatgen et al., 2009) Key Assumptions: Serial consolidation Consolidation time: 250 ms per item Requires intermediate buffer for retaining items before they can be consolidated: Buffer not specified in model.
2. Other theories of extended sparing in RSVP Batch Consolidation (Jolicoeur et al., 2002; Raffone et al., 2015) Key Assumptions: Consolidation begins after target sequence ends Targets consolidated in a single batch Requires intermediate buffer for retaining items before they can be consolidated: Attention-based activation (Raffone et al., 2015).
2. Other theories of extended sparing in RSVP Rapid Consolidation (Gegenfurtner & Sperling, 1993; Vogel et al., 2006) Key Assumptions: Consolidation is interrupted by a mask. Thus, ability to report masked item is taken to reflect the fact that consolidation was completed before the mask. Masking experiments suggest that consolidation of a single letter can be completed in 30-50 ms.
2. Overview of theories of extended sparing in RSVP
3. A qualitative test of the theories All items lost Graded retroactive interference Graded retroactive interference All items unaffected Research Question: Can consolidation be interrupted by some kind of post-sequence event (suffix)?
3. A qualitative test of the theories Exp 1. Effect of post-sequence mask Exp 2. Effect of post-sequence simple RT task Exp 3. Effect of post-sequence 2-AFC task (parity)
Results Experiments 1-3
Preliminary Conclusion: Graded Retroactive Interference
Conditional analysis to distinguish serial vs. parallel: Does interference for L3 & L4 depend on recall of L1 & L2? It should! It should not!
Conditional analysis to distinguish serial vs. parallel: Does interference for L3 & L4 depend on recall of L1 & L2? Data Used for Analysis: 3 other experiments (Total N = 37) using 4-letter RSVP sequences, with / without post-sequence 2-AFC (auditory or visual parity judgment) Result: No difference in retroactive interference for L3 and L4 depending on whether L1 and L2 were correctly recalled: BF01 = 6.8
Conditional analysis to distinguish serial vs. parallel: Does interference for L3 & L4 depend on recall of L1 & L2?
Interpretation Interruption of WMC Task-switch effect: Response selection 2-AFC interrupts consolidation Cortical Router Model Zylberberg, Dehaene et al. 2010
In this talk: Phenomenon: Extended sparing in RSVP 2. Theories of extended sparing in RSVP A qualitative test of the theories Conclusions: WMC can be interrupted by a switch in central processing requirements, but not by a mask or another to-be-encoded item WMC occurs in parallel for successive items in RSVP WMC takes about 400 ms for a single letter in RSVP Occurrence of interruption of WMC depends on the likelihood of interference during WMC.
The more general point (perhaps more important too) Retroactive interference conflicts with typical findings of proactive interference (e.g., AB, PRP). When does one get retroactive interference and when does one get proactive interference?
“Cognitive Backward Masking” RI with Simultaneous Presentation Nieuwenstein & Wyble (JEPG, 2014) “Cognitive Backward Masking”
RI with Simultaneous Presentation Nieuwenstein & Wyble (JEPG, 2014) SOA (ms)
Little if any proactive interference! SOA (ms) Nieuwenstein & Wyble (JEPG, 2014)
Effect of T2 Probability: 100 vs. 75 vs. 50% T2 Present One would preferably contrast T2 probability within subjects but…. Block 1 = 50%, Block 2 = 100% -> Net T2 prob. block 2 = 75% Block 1 = 100%, Block 2 = 50% -> Net T2 prob. Block 2 = 75%
Interpretation of Effect of T2 Probability When T2 probability is high, the risk of interference with T1 consolidation is high. This leads to a strategy in which T1 consolidation is protected by suppressing attention for new inputs. As a result, T1 consolidation is shielded from interference but T2 processing is postponed: proactive interference (the PRP effect).
Interpretation of Effect of T2 Probability When T2 probability is low, the risk of interference with T1 consolidation is low. This leads to a strategy in which T1 consolidation does not suppress attention for new inputs. As a result, T1 consolidation is not protected, thus allowing T2 to be processed unabated, at the cost of interrupting T1 consolidation: retroactive interference.
Thank you for your attention! Working Memory Consolidation (WMC) Thank you for your attention!
Results Experiments 1-3: RTs for simple RT and 2-AFC
Exp 4. Independent Effects of 2-AFC and Masking “Cognitive Backward Masking”
Serial Position in RSVP Exp 4. Independent Effects of 2-AFC and Masking “Cognitive Backward Masking” A Accuracy B Interference 100 50 Blank + Parity judgment 90 Mask + Parity judgment 40 80 Blank - No parity judgment 70 Mask – No parity judgment 30 60 % Correct Recall % Difference 50 20 40 10 30 Blank - No parity judgment Mask - No parity judgment 20 Blank + Parity judgment 10 Mask + Parity judgment -10 1 2 3 4 1 2 3 4 Serial Position in RSVP
Conclusion Exp 4. No interaction between effects of a mask and an 2-AFC task. Mask erases sensory trace of last letter, but it does not affect consolidation of earlier letters. 2-AFC interrupts consolidation, also for letters that have already been masked by subsequent letters.