Ben Cipollini & Garrison Cottrell

Transcription

1 Ben Cipollini & Garrison Cottrell NCPW 2014 Lancashire, UK

2 A developmental approach to interhemispheric communication Ben Cipollini & Garrison Cottrell NCPW 2014 Lancashire, UK

3 Lateralization Fundamental to being Human Manual skill Language Perception 3

4 The corpus callosum intertwined with lateralization? Core research question: Does the corpus callosum cause lateralization? Hofer et al (2008)

5 The corpus callosum intertwined with lateralization? Core research question: Does the corpus callosum cause lateralization? Hofer et al (2008)

6 Origins of lateralization One big idea: hemispheric independence 5

7 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: 5

8 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! 5

9 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! More Less Brain size (g) 5

10 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication More Less callosal communication Brain size (g) 5

11 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication Less communication => greater lateralization More lateralization Less callosal communication Brain size (g) 5

12 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication Less communication => greater lateralization 6

13 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication smaller brain (macaque) Less communication => greater lateralization What about bigger brains causes this? Longer conduction delays (Ringo et al, 1994) Smaller % callosal connections (Rilling & Insel, 1999) larger brain (human) 6

14 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication Less communication => greater lateralization 7

15 Origins of lateralization One big idea: hemispheric independence Hemispheric Independence theory: Brain size matters! Bigger brains => less callosal communication smaller brain (macaque) Less communication => greater lateralization What about bigger brains causes this? Longer conduction delays (Ringo et al, 1994) Smaller % callosal connections (Rilling & Insel, 1999) larger brain (human) 7

16 Independence, you say? Don t believe the hype. 8

17 Independence, you say? Don t believe the hype. Asymmetries exist in large brains... small brains brains with no cortex (frogs, fish) brains with no spinal cord (bees) (Rogers, 2009). Habenular nucleus (frog) 8

18 Independence, you say? Don t believe the hype. Asymmetries exist in large brains... small brains brains with no cortex (frogs, fish) brains with no spinal cord (bees) (Rogers, 2009). Humans born without a corpus callosum (the fewest / slowest connections) don t show greater lateralization (e.g. Gazzaniga, 2000; Paul et al, 2007). Paul (2011) 8

19 Independence, you say? Don t believe the hype. Asymmetries exist in large brains... small brains brains with no cortex (frogs, fish) brains with no spinal cord (bees) (Rogers, 2009). Humans born without a corpus callosum (the fewest / slowest connections) don t show greater lateralization (e.g. Gazzaniga, 2000; Paul et al, 2007). There is little evidence of hemispheric independence in typically-developed humans (e.g. Stark et al., 2008). Stark et al. (2008) 8

20 Independence, you say? Don t believe the hype. Asymmetries exist in large brains... small brains brains with no cortex (frogs, fish) brains with no spinal cord (bees) (Rogers, 2009). Humans born without a corpus callosum (the fewest / slowest connections) don t show greater lateralization (e.g. Gazzaniga, 2000; Paul et al, 2007). There is little evidence of hemispheric independence in typically-developed humans (e.g. Stark et al., 2008). Interhemispheric communication increases over typical development; so does lateralization (e.g. Yakovlev & Lecours, 1967; Dundas et al., 2012; Petitto et al., 2012) Petitto et al. (2012) 8

21 Conduction Delay Magnitude does not cause interhemispheric independence, (let alone lateralization) 9

22 Ringo et al. (1994) LH Output RH Input Adapted from Ringo et al. (1994) 10

23 Ringo et al. (1994) Facts: Large brains have long axons long delays Callosal axons are particularly long. Humans have big brains that are highly lateralized. LH Output RH Input Adapted from Ringo et al. (1994) 10

24 Ringo et al. (1994) Facts: Large brains have long axons long delays Callosal axons are particularly long. Humans have big brains that are highly lateralized. LH Output RH Ringo et al. hypothesis: Longer callosal delays less interhemispheric interaction Input Adapted from Ringo et al. (1994) 10

25 Ringo et al. (1994) Facts: Large brains have long axons long delays Callosal axons are particularly long. Humans have big brains that are highly lateralized. LH Output RH Ringo et al. hypothesis: Longer callosal delays less interhemispheric interaction Ringo et al. assumptions: Less interhemispheric interaction greater lateralization Input Adapted from Ringo et al. (1994) 10

26 Ringo et al. (1994) model Output t=0 LH RH All intrahemispheric connection delays = 1 time-step Input Adapted from Ringo et al. (1994) 11

27 Ringo et al. (1994) model Output t=1 LH RH activation +1-1 Input Adapted from Ringo et al. (1994) 12

28 Ringo et al. (1994) model Output t=5 LH RH activation +1-1 Input Adapted from Ringo et al. (1994) 13

29 Ringo et al. (1994) model Output t=10 LH RH activation +1-1 Input Adapted from Ringo et al. (1994) 14

30 Ringo et al. (1994) model Output t=15 LH RH activation +1-1 Input Adapted from Ringo et al. (1994) 15

31 Ringo et al. (1994) model Output t=tend=25 LH RH activation +1-1 Input Adapted from Ringo et al. (1994) 16

32 Ringo et al. (1994) methods LH RH Adapted from Ringo et al. (1994) 17

33 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: LH RH Adapted from Ringo et al. (1994) 17

34 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) LH RH Adapted from Ringo et al. (1994) 17

35 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) Remove the corpus callosum LH X RH X Adapted from Ringo et al. (1994) 17

36 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) Remove the corpus callosum Compute the performance decrease LH X RH X Adapted from Ringo et al. (1994) 17

37 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) Remove the corpus callosum Compute the performance decrease Perf. decrease degree of communication LH RH Adapted from Ringo et al. (1994) 17

38 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) Remove the corpus callosum Compute the performance decrease Perf. decrease degree of communication Small decrease: little communication Big decrease: lots of communication LH RH Adapted from Ringo et al. (1994) 17

39 Ringo et al. (1994) methods Measure of inter-hemispheric dependence: Measure current performance (100% post-training) Remove the corpus callosum Compute the performance decrease Perf. decrease degree of communication Small decrease: little communication Big decrease: lots of communication LH RH Manipulations: Interhemispheric delay: 1 or 10 time-steps Tend: varied between 15 and 75 total time steps Adapted from Ringo et al. (1994) 17

40 X Ringo et al. (1994) results X % correct Tend 18

41 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Tend 18

42 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Delay=1 starts out poorly and drops to chance quickly. Tend 18

43 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Delay=1 starts out poorly and drops to chance quickly. Tend How much is explained by the difference in onset? 18

44 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Delay=1 starts out poorly and drops to chance quickly. Difference is explained completely by 10-1=9 time-step difference to onset. Tend 19

45 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Delay=1 starts out poorly and drops to chance quickly. Difference is explained completely by 10-1=9 time-step difference to onset. Tend 19

46 X Ringo et al. (1994) results X % correct Delay=10 initially has good performance, quickly drops off. Delay=1 starts out poorly and drops to chance quickly. Difference is explained completely by 10-1=9 time-step difference to onset. Tend All of it! The hemispheres coordinate exactly the same, just with a delay to the onset. 19

47 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay 20

48 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric communication is precisely the same between short and long delays, except for the timing to onset. 20

49 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric communication is precisely the same between short and long delays, except for the timing to onset. Does an onset delay matter? In most cases, probably not. 20

50 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric communication is precisely the same between short and long delays, except for the timing to onset. Does an onset delay matter? In most cases, probably not. The delay differences are very small much smaller than those used in the model Delay (µs) 20 Adapted from Caminiti et al. (2009)

51 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric communication is precisely the same between short and long delays, except for the timing to onset. Does an onset delay matter? In most cases, probably not. The delay differences are very small much smaller than those used in the model. There are many delays between the hemispheres and many are very short Delay (µs) 20 Adapted from Caminiti et al. (2009)

52 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric communication is precisely the same between short and long delays, except for the timing to onset. Does an onset delay matter? In most cases, probably not. The delay differences are very small much smaller than those used in the model. Anterior Posterior There are many delays between the hemispheres and many are very short. The left hemisphere could predict the right hemisphere s activity (5-20ms). Adapted from Aboitiz et al (1992) 20

53 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric Take-aways: communication is precisely the same between short Conduction and long delays, delays except only for affect the timing the to onset. (and not the quality) of communication Does an onset delay matter? In most cases, probably not. The delay differences are very small much smaller than those used in the model. Anterior Posterior There are many delays between the hemispheres and many are very short. The left hemisphere could predict the right hemisphere s activity (5-20ms). Adapted from Aboitiz et al (1992) 20

54 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric Take-aways: communication is precisely the same between short Conduction and long delays, delays except only for affect the timing the to onset. (and not the quality) of communication Does an onset delay matter? In most cases, probably not. The effect may not matter at all. If it does the maximum effect is the difference in delay on a single pass. The delay differences are very small much smaller than those used in the model. Anterior Posterior There are many delays between the hemispheres and many are very short. The left hemisphere could predict the right hemisphere s activity (5-20ms). Adapted from Aboitiz et al (1992) 20

55 Ringo et al. (1994) Summary Only a short, often irrelevant onset delay In this model, hemispheric Take-aways: communication is precisely the same between short Conduction and long delays, delays except only for affect the timing the to onset. (and not the quality) of communication Does an onset delay matter? In most cases, probably not. The effect may not matter at all. If it does the maximum effect is the difference in delay on a single pass. The delay differences are very small much smaller than those used in the model. Anterior Posterior There are many delays between the hemispheres and many are very short. That s generally pretty damn The left hemisphere could predict the right small hemisphere s activity (5-20ms). Adapted from Aboitiz et al (1992) 20

56 Conduction Delay Variability does cause interhemispheric independence 21

57 Our modeling approach Developmental Model the developmental trajectory of the corpus callosum, examine interhemispheric independence. Goal: build a developmental model framework that we can later use to examine lateralization and interactions with asymmetry. Three experiments: 1. Does callosal delay variability reduce interhemispheric communication? 2. Is this a true deficit, or is it accountable by conduction delay magnitude? 3. Does developmental reduction of variability allow interhemispheric circuits to develop? 22

58 Development & delays Thin fibers conduct information with unreliable delays Axon Diameter Distribution (Cat) Unmyelinated Myelinated Post-natal day 150 Post-natal day 26 Prenatal 23 Adapted from Berbel & Innocenti (1988)

59 Development & delays Thin fibers conduct information with unreliable delays Axon Diameter Distribution (Cat) Unmyelinated Myelinated Thin, unmyelinated axons have unreliable delays (in red) (Faisal et al., 2008) Post-natal day 150 Post-natal day 26 Up to 20% variability (delay-dependent noise) (Wang, 2008) Prenatal 24 Adapted from Berbel & Innocenti (1988)

60 Humans are special in development The need for brain compression 25

61 Humans are special in development The need for brain compression Delay reliability is worst when fibers are long, thin, and unmyelinated 25

62 Humans are special in development The need for brain compression Delay reliability is worst when fibers are long, thin, and unmyelinated We predict that all three affect human infants: 25

63 Humans are special in development The need for brain compression Delay reliability is worst when fibers are long, thin, and unmyelinated We predict that all three affect human infants: Long fibers at birth, necessary to maximize adult brain size 25

64 Humans are special in development The need for brain compression Delay reliability is worst when fibers are long, thin, and unmyelinated We predict that all three affect human infants: Long fibers at birth, necessary to maximize adult brain size Thin fibers help minimize prenatal cerebral volume ( compression ) 25

65 Humans are special in development The need for brain compression Delay reliability is worst when fibers are long, thin, and unmyelinated We predict that all three affect human infants: Long fibers at birth, necessary to maximize adult brain size Thin fibers help minimize prenatal cerebral volume ( compression ) Unmyelinated fibers also minimize volume while maximizing neural plasticity. 25

66 Humans are special in development The need for brain compression 26

67 Humans are special in development The need for brain compression Hypothesis: Humans have competing goals: maximize adult brain size while having to minimize head size at birth (to reduce riskiness of birth). 26

68 Humans are special in development The need for brain compression Hypothesis: Humans have competing goals: maximize adult brain size while having to minimize head size at birth (to reduce riskiness of birth). This causes a need to maximally compress the brain at birth => best accomplished through minimizing white matter axon thickness and myelination. 26

69 Humans are special in development The need for brain compression Hypothesis: Humans have competing goals: maximize adult brain size while having to minimize head size at birth (to reduce riskiness of birth). This causes a need to maximally compress the brain at birth => best accomplished through minimizing white matter axon thickness and myelination. This is unique to humans, due to the constraint to birth a large-brained infant through a narrow, bipedal pelvis. No other large-brained animal is head-constrained at birth; human infant white matter connections should be uniquely variable in conduction delay. 26

70 Humans are special in development The need for brain compression Hypothesis: Humans have competing goals: maximize adult brain size while having to minimize head size at birth (to reduce riskiness of birth). Prenatal Postnatal This causes a need to maximally compress the brain at birth => best accomplished through minimizing white matter axon thickness and myelination. Macaque This is unique to humans, due to the constraint to birth a large-brained infant through a narrow, bipedal pelvis. No other large-brained animal is head-constrained at birth; human infant white matter connections should be uniquely variable in conduction delay. Human 26 Martin (1983)

71 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except: Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 27

72 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except:? Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 28

73 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except:? Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 28

74 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except:? Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 28

75 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except:? Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 28

76 Lewis & Elman (2008) Methods Model Architecture & Task Similar to Ringo et al., except: Only one hidden unit level Separation of inputs into left and right allows: Varying whether patterns require interhemispheric transfer or not. Lewis & Elman (2008) 29

77 Experiment 1 Methods Does noise reduce communication? Similar to Ringo et al., except we: Fix callosal delay to 10 time-steps Fix T end to give sufficient time for callosal transfer. Implement unreliability as activation noise, dependent on delay (0.1% * delay) Longer, slower fibers -> greater noise Compare no-noise (control) and noise (experimental) conditions over training epochs. 30 Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1, cc delays=10, T end =35

78 Expt 1: Constant noise Unreliability causes independence Constant noise: 31

79 Expt 1: Constant noise Unreliability causes independence Constant noise: 31

80 Expt 1: Constant noise Unreliability causes independence Constant noise: 31

81 Expt 1: Constant noise Unreliability causes independence Constant noise: 31

82 Expt 1: Constant noise Unreliability causes independence Constant noise: 31

83 Expt 1: Constant noise Unreliability causes independence Constant noise: Noise-induced independence 31

84 Expt 1: Constant noise Unreliability causes independence Raw error Lesion-induced error 32

85 Expt 1: Constant noise Unreliability causes independence Raw error Lesion-induced error intra patterns (can be done independently) inter patterns (require callosal communication) 33

86 Expt 1: Constant noise Unreliability causes independence Raw error Lesion-induced error intra patterns (can be done independently) inter patterns (require callosal communication) 33

87 Expt 1: Summary Unreliability separates the hemispheres Unreliability decreases interhemispheric dependence The effect is driven largely by inter patterns (requiring interhemispheric integration). Suggests that early in development, there is less interhemispheric communication. 34

88 Experiment 2 Methods Is the effect different from Ringo s delay onset effect? Like Ringo et al., we: Vary callosal delay [1 or 10 time-steps] Vary T end. Unlike Ringo et al., we: Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1 35

89 Experiment 2 Methods Is the effect different from Ringo s delay onset effect? Like Ringo et al., we: Vary callosal delay [1 or 10 time-steps] Vary T end. Unlike Ringo et al., we: Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1 35

90 Experiment 2 Methods Is the effect different from Ringo s delay onset effect? Like Ringo et al., we: Vary callosal delay [1 or 10 time-steps] Vary T end. Unlike Ringo et al., we: Prediction: The decrease in interhemispheric dependence is separate from the delay-to-onset found by Ringo et. al (1994). Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Why? One is caused by delay magnitude, the other by delay variability. Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1 35

91 Expt 2: intra-hemispheric patterns No effect of unreliability. Ringo et al. (1994) 36

92 Expt 2: intra-hemispheric patterns No effect of unreliability. Ringo et al. (1994) no noise 36

93 Expt 2: intra-hemispheric patterns No effect of unreliability. Ringo et al. (1994) noise = 1% 36

94 Expt 2: inter-hemispheric patterns Unreliability separates the hemispheres Ringo et al. (1994) no noise 37

95 Expt 2: inter-hemispheric patterns Unreliability separates the hemispheres Ringo et al. (1994) noise = 1% (training error) 38

96 Expt 2: Summary Unreliability separates the hemispheres This effect is different from the Ringo onset effect; this is true change in interhemispheric interactions. Again suggests greater unilateral integration early in development. No noise Noise intra-hemispheric patterns inter-hemispheric patterns Onset delay Onset delay Onset delay Onset delay + independence 39

97 Like Experiment 1, we: Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Unlike Experiment 1, we: Experiment 3 Methods Does the model show robust adult communication? Decrease variability over training Train for a greater number of epochs Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1, cc delays=10, T end =35 40

98 Like Experiment 1, we: Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Unlike Experiment 1, we: Experiment 3 Methods Does the model show robust adult communication? Decrease variability over training Train for a greater number of epochs Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1, cc delays=10, T end =35 40

99 Like Experiment 1, we: Implement unreliability as activityand delay-dependent noise (more activity, longer fiber greater noise) Compare no-noise (control) and noise (experimental) conditions. Unlike Experiment 1, we: Experiment 3 Methods Does the model show robust adult communication? Decrease variability over training Prediction: As reliability increases, the bias towards local (intrahemispheric) networks will disappear. Why? Delay magnitude does not cause independence, only delay variability does. Train for a greater number of epochs Lewis & Elman (2008) Model params: noise = 1% (mean) all delays=1, cc delays=10, T end =35 40

100 Expt 3: Simulate maturation Decreasing noise ( development ) Decreasing noise: 41

101 Expt 3: Simulate maturation Decreasing noise ( development ) Decreasing noise: 41

102 Expt 3: Simulate maturation Decreasing noise ( development ) Decreasing noise: 41

103 Expt 3: Simulate maturation Decreasing noise ( development ) Decreasing noise: Independence 41

104 Expt 3: Simulate maturation Decreasing noise ( development ) Hemispheres initially more independent, interhemispheric cooperation matures Independence Collaboration 41

105 Expt 3: Summary Reliability allows coordination between the hemispheres Increase in reliability over development allows interhemispheric circuits to develop. Developed model shows no discernible difference from the nonoise model after development. The model can account for strong interhemispheric transfer in. 42

106 Expt 3: Summary Neuroconstructivism Methods usually reduce functional circuitry (lesion, noise, pruning). This may be used as a neurally plausible method for adding functional circuitry over development. Noise disables circuits early in development Decreased noise brings those circuits online Delays don t have to be modeled to do this. Very much in-line with changes in resting state over development and change in bias from local circuits to full-brain, mature circuits. 43

107 From Independence to Lateralization Hope to see you in Pasadena! We are running this now, applying analysis techniques that allow us to assess asymmetry in representation and how that unfolds over time (single presentation) and development. Stay tuned 44

108 Summary Conduction delay magnitude: delays the onset of (not decreases), interhemispheric communication. does not have any known relationship with lateralization. Conduction delay reliability: can decrease the use of interhemispheric connections. is consistent with a developmental shift from a bias for short-range connections to long-range connections. may be used as a new tool in the neuroconstructivist s toolbox. 45

109 Acknowledgements Garrison Cottrell Semendeferi Lab Our lab (GURU) Kari Hanson Benasich Lab Gabriella Musacchia Perceptual Expertise Network (PEN) Marlene Behrmann David Plaut 46

110 Thank you! Garrison Cottrell Semendeferi Lab Our lab (GURU) Kari Hanson Benasich Lab Gabriella Musacchia Perceptual Expertise Network (PEN) Marlene Behrmann David Plaut 47

111 48

112 Expt 4: Unreliability & lateralization Interaction with task asymmetries symmetric Outputs asymmetric Inputs asymmetric symmetric 49

113 Expt 4: Unreliability & lateralization Interaction with task asymmetries symmetric Outputs asymmetric Inputs asymmetric symmetric More input features help learning 49

114 Expt 4: Unreliability & lateralization Interaction with task asymmetries symmetric Outputs asymmetric Inputs asymmetric symmetric Competing outputs impair learning 49

115 Expt 4: Unreliability & lateralization Interaction with task asymmetries symmetric Outputs asymmetric Inputs asymmetric symmetric 49

116 Expt 4: Unreliability & lateralization Interaction with task asymmetries Unreliability hurts most with asymmetric inputs (can t take full advantage of contralateral features) Unreliability helps with asymmetric outputs (partially reduces interference; helps uncover shared features) 50

117 X Ringo et al. results X Original Adapted from Ringo et al. (1994) 51

118 X Ringo et al. results X Inference: Thus, where the networks were allowed sufficient time to use the interhemispheric connections for multiple transfers of activation, a useful integration of the two hemispheres occurred. The network not allowed sufficient time for multiple transfers of activations ends up with two hemispheres operating fairly independently. (Ringo et al., 1994) Original Adapted from Ringo et al. (1994) 51

119 X Ringo et al. results X Others: This is consistent with the findings that increased brain size predicts decreased interhemispheric connectivity (Baron-Cohen et al, Science, 2001) This complementary division of processing between the hemispheres is an efficient use of cortical space because it reduces... neuronal conduction time, as short (e.g., within-hemisphere) connections are more efficient than long. (Roser, Fugelsang, Dunbar, Corballis, & Gazzaniga) Original Adapted from Ringo et al. (1994) 51

120 X Ringo et al. results X Original Adapted from Ringo et al. (1994) 51

121 Extra Extra Slides 52

122 Over development: Noise reduces collaboration No-noise Noise (1% of activity) 53

123 = + All patterns intra patterns inter patterns 54

124 Rilling data 55

125 56