Switch costs in working memory updating A tale of two foci

1
Switch costs in working memory updatingA tale
of two foci?

• Ullrich Ecker1, Stephan Lewandowsky1, Klaus
  Oberauer2
• 1University of Western Australia, 2University of
  Bristol (now University of Zurich)

2
Introduction (short)

• WM is important!
• Non-static environment requires WM updating

3
Preface Updating Components and their Costs
(Exp. 1 Ecker, Lewandowsky, Oberauer, Chee,
accepted pending minor revision JEPLMC)

• memorise (2 sec encoding time)

4
Preface Updating Components and their Costs
(Exp. 1)

• type H, remember V H A

?2
5
Preface Updating Components and their Costs
(Exp. 1)

• type A, keep remembering V H A

?
6
Preface Updating Components and their Costs
(Exp. 1)

• type M, remember M H A

M
7
Retrieval and Substitution Costs

• RT cost of substitution 300 ms
• RT cost of retrieval None.
• (i.e., ?2 F2)
• (of course also effects of the transformation)

8
Oberauers WM model (2002)
9
Question

• Null-effect of Retrieval suggests that
• Retrieval is obligatory even if not actually
  required by updating task (?2 F2).
• WM direct-access region can hold (at least) 3
  items concurrently.
• Focussing on frame brings whole bound
  representation (item-in-frame) into WMs focus of
  attention (frame switch on every updating step).
• What are the necessary conditions for obligatory
  retrieval?

10
Implicit operationalization of Substitution factor

• In first experiment, people never really knew up
  front what to do
• i.e., whether they would need access to WM
  representations
• In particular, substitutions only became clear
  after the transformation

11
Preface Updating Components and their Costs
(Exp. 1)

• memorise (2 sec encoding time)

12
Preface Updating Components and their Costs
(Exp. 1)

• type W, remember W F A

U2
13
Preface Updating Components and their Costs
(Exp. 1)

• memorise (2 sec encoding time)

14
Preface Updating Components and their Costs
(Exp. 1)

• type V, keep remembering V F A

T2
15
Aim of Exp. 2

• Isolate examine the effects of retrieval and
  substitution with a more explicit process
  operationalization

16
Exp. 2 Updating task Trial structure

• Encoding
• Remember 3 letters-in-frames
• Updating
• 7-10 steps, constant stopping probability
• Update individual frames (alphabet arithmetic
  transformation)
• Remember and type result (? RT, accuracy)
• Frame switch on every step
• Followed by free recall (not analysed)

17
Exp. 2 Design

• Updating
• 2 factors Retrieval, Substitution
• fully crossed within-subjects

18
Retrieval
encode memorise
19
R
type W, remember W F A
20
R-
type W, remember W F A
21
Substitution
encode memorise
22
S
type W, remember W F A
23
S-
type W, remember V F A
24
Putting it all together
25
R-S-
encode memorise
26
R-S-
type L, remember V F A
27
Putting it all together
28
RS-
encode memorise
29
RS-
type W, remember V F A
30
Putting it all together
31
R-S
encode memorise
32
R-S
type L, remember L F A
33
Putting it all together
34
RS
encode memorise
35
RS
type W, remember W F A
36
Exp. 2 Substitution x Retrieval
Just doing alphabet arithmetic without memory
access is quickest
37
Exp. 2 Substitution x Retrieval
Memory access takes extra timebut its the same
extra for retrieval and substitution
38
Exp. 2 Substitution x Retrieval
Doing two things to memory (S R) takes no extra
time
39
Theory

• Hypothesis
• If subjects know from the outset that they do not
  need to touch their current WM content (S-R-),
  they can perform transformations without engaging
  WMs Focus of Attention
• This should be detectable via frame switch costs
• Conditions that require the FoA
• vs. the S-R- condition

40
Perhaps

• Perhaps the Focus of Attention should actually
  be called Focus of Working Memory
• Perhaps the FoWM could be distinguishable from
  the focus of spatial attention (FoSA) in
  perception
• The FoSA clearly needed to do a transformation
  (even S-R-)
• Can the 2 foci be distinguished?

41
Exp. 3 Triplet Structure
STEP 1 (S and/or R) STEP
2 (S-R-) STEP 3 (S and/or R)
42
STEP 1 (S and/or R) STEP
2 (S-R-) STEP 3 (S and/or R)
Pure FoSA theory prediction
43
STEP 1 (S and/or R) STEP
2 (S-R-) STEP 3 (S and/or R)
.50 .50




.33 .33 .33
.33 .33 .33
1 2 3
4 5 6
Pure FoWM theory prediction
44
(No Transcript)
45
Exp. 3 Regression analysis

• It seems neither a pure FoSA nor a pure FoWM
  theory can explain the data
• Multilevel regression on logRT data indicates
  that both factors are significant
• Fixed effects            Estimate Std.
  Error t value(Intercept)  1.00560   
  0.02174   46.25Retrieval    0.20392   
  0.02484    8.21 Substitution
  0.11609    0.02376    4.89 FoSA Switch 
  0.09105    0.01576    5.78 FoWM Switch 
  0.08823    0.02703    3.26R x S      
  -0.10484    0.03467   -3.02 FoSA x FoWM 
  -0.09041    0.02887   -3.13
• So moving either the FoSA or the FoWM takes time
  (250 ms), but moving both together doesnt take
  additional time.

46
Now

• If the Focus of WM idea is correct, then
  condition 2 should be fast if step 2 involves
  either S or R. 

47
STEP 1 (S and/or R) STEP
2 (S-R-) STEP 3 (S and/or R)
.50 .50




.33 .33 .33
.33 .33 .33
1 2 3
4 5 6
48
STEP 1 (S and/or R) STEP 2 (S and/or
R) STEP 3 (S and/or R)
?B
?B
gt



?A
?A

1 2 3
4 5 6
49
STEP 1 (S and/or R) STEP 2 (S and/or
R) STEP 3 (S and/or R)
?B
?B
gt



?A
?A

1 2 3
4 5 6
50
(No Transcript)
51
Conclusions

• Transformations can be done without engaging the
  Focus of WM.
• When there is no need to touch information in
  WM
• Inconsistent with leading WM models (where the
  FoWM selects item for processing)
• It is therefore reasonable to discriminate
  between the Focus of WM and the Focus of Spatial
  Attention.

52
Thank you!
Steve Lewandowsky
Klaus Oberauer
53
Questions? Comments?
Danny
Toby
54
Exp. 3 Substitution x Retrieval
55
Exp. 3 Regression analysis

• Retrieval cost inconsistent with first
  experiment?
• Fixed effects
• Estimate Std. Error t value 
• (Intercept)  1.18353    0.02603   45.46
• nuthin      -0.17294    0.01713  -10.10
• FoSA Switch  0.08903    0.01447    6.15
• FoWM Switch  0.08465    0.02745    3.08
• FoSA x FoWM -0.09136    0.02889   -3.16
• not really Subjects in previous experiment
  always engaged their FoWM, so never did nuthin