駒場集中講義「意識の神経科学」ハンドアウトです。 1月10日(木)3・4・5限(13:00-18:00), 1月11日(金)3・4限(13:00-16:20)、講義室は駒場キャンパス15号館4階409号室。

1. 広域システム科学特殊講義Ⅴ 「意識の神経科学」ハンドアウト
1月10日(木)3・4・5限(13:00-18:00), 1月11日(金)3・4限(13:00-16:20)
駒場キャンパス15号館4階409号室
吉田正俊 (生理学研究所・認知行動発達部門・助教)
日本語要旨 (To-be-updated): 1) 意識のうち、報告可能性などの心理学的側面である気づ
き(awareness)の神経相関を見つける仕事が行われた。 このような研究では視覚入力はまった
く同じであるにも関わらず、それが見えたかどうかによって応答が変化する神経活動が同定
されてきた。盲視は視覚経験がないのにも関わらず視覚情報処理が可能な症例である。盲視
モデル動物でも検出報告に対応する神経活動が見つかる。つまり、検出報告は気づきと同一
化できない。 このことはimplicit perceptionにおけるdirect measureとindirect measureの成
2)
績の乖離から考えても納得がいく。気づきを定量化する方法として信号検出理論について解
説した。信号検出理論では、盲視はyes-no detectionとtwo-alternative forced choiceという二
つのdirect measureの成績が乖離するものであると説明することが出来る。3) 気づきと注意
とは非常に関連した現象である。近年、両者を単離する試みが行われて来ており、気づきと
注意とは脳内でもべつの部分によって処理されるべつの過程であることが明らかになってき
た。ボトムアップ性注意の計算論モデルである「サリエンシー・マップ」によって盲視サル
での視覚探索能力が明らかになった。このことから盲視では意識的な視覚情報処理とはべつ
にサリエンシーによる意識下の情報処理があることが明らかになった。サリエンシー処理系
が盲視における「なにかあるかんじ」に寄与しているのかもしれない。4) 両側LOの障害に
よってvisual agnosiaになった患者DFの研究から、視覚情報処理は腹側視覚経路が意識的な
認知のための知覚に、背側視覚経路が無意識的な行動のための知覚に関わることが提唱され
た。この文脈で盲視はV1の損傷によって皮質下経由で視覚情報が背側視覚経路のみにアクセ
5)
ス可能になった状態と解釈することが出来る。 知覚に運動が必要であることは網膜像の固
定実験などから提唱されてきた。 また、 眼球運動の情報がefference copyとして視覚応答に影
響を与えるなどの予測的情報処理が行われていることが明らかになっている。また、計算論
的モデルなどから、視覚ニューロンが外界からの入力のフィルターとして受動的に反応する
のではなくて、トップダウンの予測からの誤差を計算していることが示唆されている。意識
の情報処理の解明のためには、背側視覚経路での外界との相互作用のループが腹側視覚経路
での内部モデルを作り上げ維持していく、といった相互作用を考える必要があることを提唱
してこの講義を終える。
• A common-sense definition of consciousness:
1. Overview, NCC and blindsight "consciousness refers to those states of
sentience or awareness that typically begin
1-1 What is consciousness? when we wake from a dreamless sleep and
• Let’s start from examples. continue through the day …”
• Salient visual stimuli can disappear. Not all of • Access-consciousness (= awareness) and
the info on the retina is conscious. Phenomenal consciousness (= qualia)
1-2 Definition of consciousness 1-3 What is awareness?
• Let’s start from a common-sense definition, not • The ability to discriminate, categorize, and
from an analytic definition. react to environmental stimuli; the integration

2. of information by a cognitive system; the • Blindsight in philosophy of mind
reportability of mental states; the focus of • Possible philosophical zombie: We can
attention; the deliberate control of behavior conceive a human who retains all of
• Definition of awareness: “a state wherein we sensorimotor ability but have no
have access to some information, and can use phenomenological consciousness at all.
that information in the control of behavior.” • David Chalmers (Conscious mind (1996),
“The psychological concept of mind" Ch.6.3): “(The description of blindsight) is
• Neural correlates of awareness: a visual input compatible with the coherence between
is constant but perception of that visual consciousness and awareness.”
stimulus varies • Ned Block (On a confusion about a function of
consciousness (1995)): “But stimuli in the blind
1-4 Neural correlates of awareness (1): field (of blindsight) are BOTH
bi-stable percept access-unconscious and phenomenally
• In binocular rivalry, the stimulus is stable but unconscious.”
the content of awareness switches. Then, we • Daniel Dennett (Consciousness explained
can find the neural correlate of visual (1991), Ch.11.2): “As we shall see, however,
awareness by comparing two different blindsight does not support the concept of a
perceptual reports. zombie; it undermines it.”
• Activity of IT neurons reflects the monkeys’ • Let’s look at real blindsight.
perceptual report1. This is strong evidence that • Although the patients have no visual
IT neurons represent content of subjective experience like ours, but they sometimes insist
experience. they had a feeling of something happening in
• IT neurons represent a stage of processing their blind field.
beyond ambiguities and their activity reflects the • It is not phenomenal vision but a kind of
brain's internal view of objects. subjective experience.
• Revised definition of blindsight: The visually
1-5 Neural correlates of awareness (2): evoked voluntary responses of patients with
near-threshold stimuli striate cortical destruction that are
• In the detection task, subjects are required to demonstrated despite a phenomenal
answer whether the sensory stimulus is blindness.
present or absent. Key comparison is Hit vs.
Miss. 1-7 Blindsight in monkey
• Strong modulation between Hit and Miss in • Blindsight occurs after 2-3 months training.
early visual cortex2 and medial premotor • Are the monkeys really ‘blind’ to the visual
cortex3. stimuli? In yes-no task, the monkeys behaved
as if it is a No-target trial. The monkeys are 'not
1-6 Blindsight able to see', as in human blindsight.
• Q: What is blindsight? A: A: "The visually • The stimulus-related response in the superior
evoked voluntary responses of patients with colliculus was modulated by Hit - Miss.
striate cortical destruction that are • V1 lesioned monkeys showed residual
demonstrated without awareness" detection.
• Phenomenal consciousness can be • Detection is not necessary reflect awareness.
dissociated from visual information processing. • Detection-related activity may be a neural

3. correlate of "feeling-of-something-happening”. • Neural correlate for Implicit perception in Left
fusiform gyrus.
1-8 Conclusion • When we take into account of implicit
• Detection is not necessary equated with perception, we cannot equate Detection with
awareness. => Signal detection theory (SDT) Awareness.
in #2 lecture Measuring consciousness” • Blindsight is different from implicit perception
• Neural correlate of awareness in hit-miss because of dissociation between two direct
comparison is not the final destination but the measures (discrimination and yes-no
starting point. detection).
2. measuring consciousness 2-2 Signal detection theory (yes-no
detection)
2-1 Implicit perception • The V1-lesioned monkeys behaved as if it is a
• What is awareness? (philosophy of mind) No-target trial.
• Definition of awareness: • The monkeys are ‘not able to see’, as in
• “a state wherein we have direct access to human blindsight.
some information, and can use that • Here the deficit in detection / awareness was
information in the control of behavior.” evaluated by the correct ratio.
• A subject is aware of some information when • This is not a good idea because the blindsight
that information is directly available to bring monkeys can simply get conservative (=>
to bear in the direction of a wide range of When in doubt, say no).
behavioral processes. • The criteria / bias affects the correct ratio.
• This allows for the possibility of experience in • What is better index for detection / awareness?
non-human animals ... => Signal detection theory (SDT)
• What does he mean by “direct”? • You have a sensor (1D continuous value). You
• Operational definition of two measures have to decide which is a signal and which is a
• Direct measure: noise, based on the sensor value. When you
• The discriminative response is part of the classify the data as signal, you are aware of
task definition. the signal.
• Direct, explicit report of the stimuli 1) You collect samples.
• Indirect measure: 2) You set the criteria for optimal discrimination.
• The discriminative response is not part of the 3) You classify new data by comparing the
task definition. sensor value and the criteria.
• Indirect report of the stimuli • It is a recognition model (model-free).
• Implicit perception is defined as Unknown processes generate data and then
• Above-chance performance in indirect they are classified with criteria (c=2).
measure • In the generative model (model-based),
• Chance-level performance in direct measure processes with unknown parameters (Noise:
• Masked repetition priming N(0,1); Signal: N(d’,1)) generate data and then
• Is the target natural or manmade? => Indirect we estimate parameter (d’ = 4) and classify
measure of prime with criteria.
• Is the prime natural or manmade? => Direct • The sensitivity of the sensor is characterized
measure of prime as d’. d’ is independent of criteria (c). (The

4. correct ratio depends on c.) the stimuli
• OK, but we have no such sensor. How to • If the subject report that they have no
estimate d’ in psychophysics? awareness, there is no awareness.
1) You set a criterion and classify samples. • pros: direct
2) You get the feedback (correct or incorrect). • cons: subjective - influenced by the subject’s
3) You obtain data set 1 (with hit, miss, FA, CR). bias
4) Repeat 1)-3) with different criteria. • Objective method:
5) You reconstruct the distribution of samples. • To evaluate subjects performance on
6) You estimate d’. detection task or discrimination task
• How do you change the criteria? • If the performance is at chance level, there is
1) Confidence rating (Human study) no awareness.
2) By changing value or probability (animal • pros: objective, operationally well-defined
study) • cons: indirect
• Blindsight in normal
2-3 Signal detection theory (forced-choice • Perception – Objective measure
discrimination) • Awareness – Subjective measure
• In 2afc, we have two sensors. Thus SDT can • Better performance in Perception than that of
be represented as in 2D. Awareness
2-4 Application of SDT to blindsight 2-7 Application of SDT to brain imaging
• d’ for Yes-no task and d’ for 2afc were • Strong modulation between Hit and Miss in
compared4. early visual cortex (V1, V2, V3).
• d’(discrimination) > d’(detection) => Objective • FA has stronger modulation than Misses. This
characterization of blindsight is consistent with SDT.
• d (detection) > 0 => Detection Awareness
• Note that both are direct measures. 2-8 Accumulator model
• Blindsight can be redefined as d'YN > d'FC • So far, we only care about the task
• Both tasks (Yes-No task(detection) and 2AFC performance (correct or incorrect). But the info
task (discrimination)) evaluates direct of reaction time is useful to estimate decision
measure. processes.
• Blindsight suggests that direct measures have • Perceptual decision making in reaction-time
different degree of conscious access or task
awareness. • LIP cells behave like accumulated evidence.
• Accumulator model: Where is awareness? At
2-5 Application of SDT to blindsight monkey which stage one becomes aware?
• Better performance in discrimination (FC) than • Extension of diffusion model: The boundary
detection (YN). => Consistent with data from (threshold) can be time-dependent.
human blindsight patient • Need for integrating SDT and diffusion model
2-6 Objective measure and subjective 2-9 Other measures of awareness
measure • Post-decision wagering objectively measures
• Subjective method: awareness5.
• To simply ask subjects whether they can see • Metacognition can be a test for awareness in

5. certain animals. awareness?
• It can reflect a process of selective attention of
3. Attention and consciousness competing objects.
• the phenomenon of binocular rivalry is also a
3-1 What is consciousness and attention? form of visual selection” “this selection occurs
• Let’s start from examples. … even in the absence of explicit instructions
• Motion-induced blindness: attention affects to attend to one stimulus or the other.”
conscious experience6. • Bi-stable stimuli + discrimination:
• Change blindness: we cannot consciously • Always accompany selection of multiple
perceive a big change in a scene when we do objects
not pay attention to it. • Confounded with selective attention
• Near-threshold stimuli + detection:
3-2 What is attention? • Always accompany trial-by-trial variation
• William James: "Everyone knows what • Confounded with sustained attention (=
attention is. It is the taking possession by the arousal)
mind in clear and vivid form, of one out of what • Using CFS, perceptual suppression is possible
seem several simultaneously possible without selection of competing stimuli.
objects...” • V1 activity reflect attention, not awareness7.
• Taxonomy of attention:
• Selective attention: ability to focus on 3-5 Computational model of bottom-up
positions or objects attention (saliency)
• Sustained attention: alertness, ability to • Dictionary definition of “salient”: “very easy to
concentrate notice”
• Source of selective attention • Saliency: related to selective attention and
• Bottom-up: stimulus-driven (pre-attentive, bottom-up attention of positions
pop-out) • Feature integration theory - parallel search
• Top-down: goal-directed (pop-out) and serial search
• Bottom-up vs. top-down attention • Original computational model
• Visual search • A unique, Saliency map, independent of
• Pre-cue task Feature maps
• When top-down attention is manipulated by • Winner-take-all rule (WTA)
pre-cue, both dorsal and ventral visual • Selection was made on the saliency maps
pathways are activated. • What is saliency map? - An explicit 2D map
that encodes the saliency of objects in the
3-3 Attention and awareness (1) visual environment. A computational concept.
• How can we dissociate attention and
awareness? - [Attended vs. unattended] * 3-6 Saliency in blindsight
[Aware vs. Unaware] • Is blindsight available in daily life?
• Attention and awareness are different entities. • Salient stimuli attract gazes of blindsight
• They are process in different brain regions. monkeys.
• Blindsight in daily life.
3-4 Attention and awareness (2) • Cue-less blindsight is a philosophical zombie,
• Does binocular rivalry reflect visual in a way.

6. • Blindsight monkeys are able to detect colors. experience, only saying that objects tend to
• The blindsight monkeys respond to color appear ‘blurred’ and that separate elements
saliency, not to color identity. ‘run into each other’.
• Single-line copying task: DF developed a
3-7 Conclusion strategy to overcome perceptual deficit by
• What it is like to be blindsight? - Dual system visuomotor skills.
for conscious vision and saliency. • Lesson learned: Perceptual task can be solved
• Feeling ‘atmosphere’ is saliency without visual by visuomotor skill. It is very important to
consciousness. control tasks not to develop unexpected
strategy.
4. Two visual systems hypothesis • One of the criticisms about two-systems
(To-be-updated) hypothesis is that the task they used is not well
controlled.
4-1 Comparative view • DF’s deficit can be explained by selective
• Phylogenically ancient brain supports deficit in visual processing in allocentric
blindsight. frame10.
• Two visual systems in rodents
• Superior colliculus for orienting 4-3 Zombie in us
• Visual cortex for pattern discrimination • Size-contrast illusions deceive the eye but not
• Two visual systems in the frog8 the hand.
• Obstacle avoidance is thalamic. • Large adjustments in visually guided reaching
• Response to preys is collicular. do not depend on vision of the hand or
• Two visual systems in birds perception of target displacement.
• Nucleus rotundus -> Entopallium as • Blindsight in normal
homologue of Pulv -> Extrastriate cortex • A direct oculomotor correlate of unconscious
visual processing11.
4-2 Visual agnosia
• Optic ataxia - Bilateral damage in dorsal visual 4-4 Blindsight in relation to two visual
pathway (Balint’s syndrome) pathways
• Visual form agnosia9 (patient DF): Bilateral • Blindsight depends on the superior colliculus.
damage in ventral visual pathway (LO) • Blindsight depends on LGN12.
• DF matched her card orientation to the slot • Possible color pathway in blindsight
during the course of the movement, well before • Retina -> SC -> Pulvinar
contacting the target. • Retina -> LGN -> extrastriate cortex
• Functional double dissociation • Possible contribution of training and plasticity
• What exactly Goodale said:
• "Visual phenomenology ... can arise only 5. enactive view / active vision
from processing in the ventral stream ..." (To-be-updated)
• "We have assumed ... that visual-processing
modules in the dorsal stream ... are not 5-1 Microsaccades
normally available to awareness." • Microsaccades occur during free-viewing
• Perception & visual experience of DF
• She has difficulty describing her visual 5-2 Remapping and saccadic suppression

7. • What is ‘corollary discharge’ / ‘efferent copy’? • A neuron has a belief (=model = prior).
• First formalized by Helmholtz (1866) • Predictive brain hypothesis
• why image motion caused by saccades • Forward prediction error = surprise =
passes unnoticed bottom-up attention
• Why stability is maintained in spite of shifts in • Backward prediction = model in the brain =
image position conscious perception
• Efferent copy (Von Holst and Mittelstaedt
1954) 5-5 Brain or environment, which determines
• Keeping track of one's movements was conscious experience?
based on monitoring outputs to muscles • Blindsight = Frog’s consciousness?
(implication of exact copy) • What is it like to be blindsighted?
• Corollary discharge (Sperry 1950) • Which is determinant of conscious experience,
• The interaction of "motor patterns" with a sensory input ? (=> Externalism, enactivism)
"sensorium" without specifying where the or brain activity? (=> Internalism)
interaction occurs • Phantom limb: the case for brain activity
• Corollary Discharge is used in many situations • Braille reading: the case for sensory input
• Why can’t you tickle yourself? - Cancellation • Inverted glass: the case for sensory input
of self-produced sensation • Rewiring thalamocortical pathway in ferret14:
• Stability of vision during saccade - the case for sensory input15
Peri-saccadic remapping • Which occurs in blindsight?
• Programming of sequential movement - A • Internalist view: ‘Feeling of something
pathway in primate brain for internal happening’ is a result of SC activity but it is
monitoring of movements13. overridden by ‘redness of red’ in normal
• Pre-saccadic remapping in LIP subjects. After V1 lesion, it was unmasked.
• Influence of the thalamus on spatial visual • Enactive view: ‘Feeling of something
processing in frontal cortex. happening’ can be a kind of conscious
experience accompanied by functional
5-3 Bayesian surprise recovery and expanded availability of
• How to evaluate temporal change as opposed sensorimotor skill.
to spatial saliency?
• In term of information-theoretical view, surprise 5-6 Let’s combine everything
can be expressed as a change from prior to • Dorsal = feeling of something = saliency
posterior. • Ventral = qualia
• You have a belief about the world: “it is the • Dorsal pathway is not enough. Sensorimotor
CNN news.” hypothesis fails.
• We get surprised if we have to change our • Ventral pathway is not enough. It is not a filter.
belief of the world. • History of establishing internal model by
• The world becomes boring again. predictive coding does matter.
• Surprise is better predictor than saliency.
文献
5-4 Bayesian surprise and predictive coding
1. Sheinberg, D. L. & Logothetis, N. K. The role
• Neuron can be a ‘surprise detector’.
of temporal cortical areas in perceptual
• V1 response can be modeled by surprise. organization. Proc Natl Acad Sci USA 94,
3408–3413 (1997).

8. 2. Ress, D. & Heeger, D. J. Neuronal correlates
of perception in early visual cortex. Nat
Neurosci 6, 414–420 (2003).
3. de Lafuente, V. & Romo, R. Neuronal
correlates of subjective sensory experience.
Nat Neurosci 8, 1698–1703 (2005).
4. Azzopardi, P. & COWEY, A. Is blindsight like
normal, near-threshold vision? Proc Natl Acad
Sci USA 94, 14190–14194 (1997).
5. Persaud, N., McLeod, P. & Cowey, A.
Post-decision wagering objectively measures
awareness. Nat Neurosci 10, 257–261 (2007).
6. Schölvinck, M. L. & Rees, G. Attentional
influences on the dynamics of motion-induced
blindness. Journal of Vision 9, 38.1–9 (2009).
7. Watanabe, M. et al. Attention But Not
Awareness Modulates the BOLD Signal in the
Human V1 During Binocular Suppression.
Science 334, 829–831 (2011).
8. Ingle, D. Two Visual Systems in the Frog.
Science 181, 1053–1055 (1973).
9. Goodale, M. A., Milner, A. D., Jakobson, L. S.
& Carey, D. P. A neurological dissociation
between perceiving objects and grasping
them. Nature 349, 154–156 (1991).
10. Schenk, T. An allocentric rather than
perceptual deficit in patient D.F. Nat Neurosci
9, 1369–1370 (2006).
11. Rothkirch, M., Stein, T., Sekutowicz, M. &
Sterzer, P. A direct oculomotor correlate of
unconscious visual processing. Curr. Biol. 22,
R514–R515 (2012).
12. Schmid, M. C. et al. Blindsight depends on the
lateral geniculate nucleus. Nature 466,
373–377 (2010).
13. Sommer, M. A. A Pathway in Primate Brain for
Internal Monitoring of Movements. Science
296, 1480–1482 (2002).
14. Sharma, J., Angelucci, A. & Sur, M. Induction
of visual orientation modules in auditory
cortex. Nature 404, 841–847 (2000).
15. Melchner, von, L., Pallas, S. L. & Sur, M.
Visual behaviour mediated by retinal
projections directed to the auditory pathway.
Nature 404, 871–876 (2000).