PS2080A:  Conceptual Issues in Psychology
Term II, FRIDAY 10 – 12 am (Room 290 Yorkon)

Lecture 5:    Neuroscience II : Current Approaches

Johannes M. Zanker,, (Room 218)

Topics Lecture 5:

Neuroscience: different levels of explanations - a variety of methods

the scale of phenomena to be studied in the nervous system ranges from less than a nonometer (molecular events) to the range of kilometers (interactions in biological social systems) - 12 orders of magnitude! this requires a wide range of different methodologies:

Neurones - making decisions

the most immediate demonstration that neurones are involved in generating behaviour would be to show that a behavioural decision is correlated with the activity of a single neuron; even better: change the activity of the neuron in question and observe the change of behaviour (Newsome et al. 1989, 1990).

the relation between the activity of a single neuron and the behaviour (reported percept) of a monkey can be directly observed by electrophysiological recording from MT during a direction discrimination task: random dots are moving to the right or to the left & monkey is pressing corresponding button; number of dots is varied to test different levels of stimulus strength

the number of correct responses about motion direction made by the individual MT neuron (quite) precisely corresponds to the behaviour (discrimination performance) of the monkey

the discrimination behaviour (percept) can be changed (biased towards the preferred direction) by stimulating the MT neuron

Neurones - representing features

another crucial property for neurones to be involved in mental events is the representation of features and events in the outside world; this is a phenomenon that is readily found in sensory system - neurones are 'tuned' to very specific types of optical information in the retinal images

from Tanaka et al. 1991

  • the receptive field structure of individual neurones generates a characteristic selectivity: 'filter' properties (tuning)
  • simple features like brightness, contrast, orientation, colour, or motion are encoded (represented) in the early visual stream
  • in ‘higher’ brain regions, we find much more specific responses of individual neurones for feature combinations : objects such as geometric figures, fruits, or hands are represented in the temporal lobe (inferotemporal cortex: IT)

Neurones - so are they incorporating mental events ?

a case in point: individual neurones were identified in area IT of the monkey cortex that respond to specific faces ('grandmother cell'?), face movements (linked to emotions!), direction of gaze (social communication), and to particular views of faces (shown here: neural activity -top- for different views -bottom-) (Perrett et al 1989)

From neurones and networks to modules

the concept that neurones are organised in functional units (modules), which are corresponding to particular brain regions, has a long tradition (see neuroscience history lecture 1)

speculative approach of phrenology (Franz Gall, 1812)  

>>   identification of about 40 different brain regions by studying cyto-architecture & functional anatomy through brain lesions (K. Brodmann, 1909)

Mapping function and space : fMRI

modern functional imaging techniques (in particular, fMRI) allow to identify brain regions that are dedicated to specific tasks - functional modules - and to study the internal organisation of these modules

brain activation in a subject performing a task of covert spatial attention

a variety of functions is mapped to the brain

functional imaging can also be used to observe the mapping of visual space on the human cortex

retinotopic maps demonstrated in the living brain (T. Morland, pers. communication)

<< this expanding pattern stimulates all regions in the visual field in a cyclic fashion

it leads to a systematic change of brain activity in the striate cortex as shown by the red marks (retinotopic map) >>

this is an in-vivo (real-time in living humans) confirmation of cortical design principles !!

Human brain activity reflecting perception

fMRI studies also demonstrate cortical activation in specific brain regions that correspond exactly to particular visual illusions (similar to the activity of a single neuron corresponding to behaviour): for instance, motion aftereffects (waterfall illusion) in area MT (Tootell 1998)


the timecourse of cortical activation in area MT after stopping an adapting motion stimulus shows a peak that correlates well with the perceived illusion srength

Consciousness revealed (?)

the holy grail of functional imaging is to discover brain activity correlated with high-level mental activity, such as consciousness

specific activation in a single region was related to ‘willed action’ (Hyder et al 1997) - still a matter of debate

activation correlated to imagination : activation without sensory input! (ffytche et al 1998; Kreiman et al 2000)

neural correlates of awareness during binocular rivalry and other bistable percepts are localised in parietal and prefrontal brain regions (Rees et al. 2001): when two different images are porjected on the two eyes (face and house, for instance), the binocular image superimposes both scenes at the same time - but the observer can perceive the two images in alternation; there are brain regions in the dorsal and frontal brain which are activated in synchrony with the perceptual switching between the two images - this is interpreted as reflecting activity related to becoming aware of one or the other image

what do these findings tell us about the mind (and body) ?

we can relate these results fromphysiological experiments to the main solutions of the body-mind problem

… think ...        (see also chapter 2 of Blackburn 1999)

Psychophysics: do we know the world ?

which of the two vertical lines is shorter?

the misjudgment of length is known as Muller-Lyer illusion

Gregory, 1968, suggested a knowledge-based interpretation of this illusion ‘inappropriate constancy scaling theory’ : in the 3D-interpretation the line that is perceived as closer is re-scaled to correct for larger angular size of closer objects

A low-level, neuroscientific explantion

measure a simple behavioural response in a simple neural system by recording the direction of flight when a fly is looking at a Muller-Lyer figure: ‘gaze control’ in flies  

the distance between the peaks in the distribution of flight direction (neuronal activation in the control system) corresponds to the apparent length experienced as Muller-Lyer illusion !!

the distribution of activation in a simple neural model (low-pass filtering of stimulus image) shows the same properties: distance between peaks (blue arrow) is smaller for inward than for outward configuration

human eye movements can show similar patterns !!
the 'neural model' demonstrates that the rules of processing themselves can distort the sensory information in its neural representation >> this suggests, for this example, that there is no veridical representation of length in the visual system !!
this simple explanation of the illusion as (mechanistic, nonveridical) property of neural representation is in accordance with the neurone doctrine (properties of neurons explain perception and behaviour)

Blindsight : seeing what can’t be seen

patient D.B. suffers from a blind region in the visual field region after cortical lesion - nevertheless D.B. is able to report blinking & motion direction of targets in this blind visual field !! (Weiskrantz 1995)
this phenomen - blindsight - is interpreted as suggesting subcortical processing pathways that lead to a cortical representation which is not available to awareness

blindsight can also be demonstrated in ‘normal’ observers as unconscious perception of particular features; for instance in binocular rivalry displays generating unaware structure (e.g. orientation or motion direction differs between the stimuli of the two eyes, in binocular view they are seen simultaneously - a region in which local image features are exchanged between the two eyes can be localised by the observer, although the image texture looks the same everywhere) (Kolb & Braun 1995)

Change blindness : missing the obvious

conversely, we can easily miss major changes in a scene (O'Regan et al 1988)

flicker paradigm: the change in this image is concealed by white frames shown in alternation with the two (different) images

mudsplash paradigm: the change in this image is concealed by the random patches that are shown once and again

once again, the visual system turns out to be rather unreliable…
such awareness phenomena are often regarded by visual scientist as indicators of general mechanisms of consciousness, and therefore closely related to the mind

So why should we trust our eye (brains)?

our sensory system provides no veridical but a pragmatic image of world - simple operating rules driven by sensory input are designed to generate appropriate behavioural responses even in the absence of veridical perception
illusions that are obvious in perception are not visible in action : grip is appropriate for physical, not perceived size
(evidence is debated by Carey 2001)
distinction between vision for perception and vision for action is reflected in the organization of the visual pathways in primate cerebral cortex : ventral and dorsal stream (Milner & Goodale)

Modelling : a variety of approaches

models are the formailsed attempt to understand the mechanisms of neural information processing; they can be found in various forms and have a long trandition

the common purpose of all models is:

Marr: levels of modelling

a major advance in the modelling of neural activity is found in the work of D. Marr (1982), who distinguished three levels of modelling

a typical (schematic) model of sentence processing, in form of box-and-arrow diagram, segments a task in logical processing steps but does not explain how the sub-tasks are performed (e.g. how does the model 'find word meaning'?)

the treatment of a modelling problem suggested by Marr provides a guideline to move ahead from vague flowcharts to exact simulations

Models of behaviour : from vehicles to robots

cybernetics / control theory : combine sensory input and interaction rules and motor outputs such as to design machines that show specific behaviour
basic principles incorporated in ‘vehicles’ (V. Braitenberg, 1984): control logic predicts behaviour
(recent goal: autonomous helicopter)

a typical experimental system is sometimes called 'turtle': it is equipped with two sensors (for instance, light sensors), two motors, and can have a variety of connectivity rules (for instance, sensory stimulation reduces motor output: inhibition)

    differential effects on left/right vehicle motor >> speed & directions changes
  • uncrossed negative feedback: ipsilateral inhibition >> ‘blind love
  • crossed negative feedback: contralateral inhibition >> ‘attentive admiration

more complex systems can be designed that mimick complex behaviour: 'synthetic psychology' (V. Braitenberg, 1984)

Khepera, K-Team SA, Switzerland: larger sets of sensors & connections generating complex behaviour (a ‘value system’)

‘autonomous’ robots can use multimodal and extended sensory systems, a variety of motors and actuators, versatile tools (like arms) & intelligent connectivity

vehicles are implementations of computational models to explain and test behavioural strategies (Webb 2002)
they are the basis of cutting-edge technologies (e.g. UAVs …)

Applied modelling : neuro-prosthesis

neuroscientific models can provide the key technology to design artificial (replacement) limbs to be used in prosthesis of the future: this model has a large set of 'muscles', 'bones' and 'joints', just modelled after a human arm

the major issue now is how to attach these artificial limbs to the living body and allow direct control from the nervous system
the key experiment as been recently published in an animal system: remote control of a robot arm from a monkey's brain

  • chronic implantation of large electrode array (96) in motor cortex of monkey
  • long-term recordings for correlational analysis of neural code of motor commands
  • transfer data through web to robotics department
  • drive robot arm with transmitted signals!!

Interface between (wo)man & machine

consider the speed of technological development, AI, increasing relevance of IT-based technologies and compare it with the speed of evolution
  • growth rates of WWW
  • acceleration of computing speed, software efficiency (neural nets, genetic algorithms)
  • novel interfaces between nervous system and electronic devices

    Moore's law predicts the processing power of man-made computers - around 2030 a PC could be as smart as a human brain

when will the engine of evolution generate secondary intelligence that will exceed the intelligence of its creators? (Kurzweil 1999)
'Robo sapiens' predicts the fusion, or at least close coexistence, of man and (intelligent) machine - an 'apocalyptic optimism'?

'building better humans'

The (classical) ‘cognitive revolution’

the basic approach of cognitive psychology is to explain how the mind works in terms of function !
the key is to identify network of essential information processing steps, and describe them with box & arrow diagrams, in analogy with software (flowchart)

issues arising from this approach:

N.B. it also should be noted that flowchart (black box) analysis is not the invention of the 'cognitive revolution' but has been introduced before: in cybernetics this is the standard technique to analyse and predict behaviour (leading to its use in softaware development)

The ‘neuroscientific proposition’

the basic approach of neuroscience is to explain how the mind works in terms of substrate (brain, neural networks) !
the key is to identify the basic physical (biological) processes (of neurones and networks) which define the processing constraints, and describe all mental operations as combination of such basic processes, in analogy with hardware (circuits)

issues arising from this approach:

A ‘synoptic view’

what about a combination of a top-down and bottom-up approach?
does an iterative & recursive analysis allow to overcome the difficulties that each approach generates when used in isolation; this is what some people understand by 'cognitive neuroscience' - the functional analysis of cognitive tasks in close relation to our knowledge of neuronal function and functional architecture of the brain!

even son, there are some more fundamental sceptical thoughts, such as: can a system understand itself?
(i.e., can a system have enough operations available to map all possible operations of this system?)

Evaluation of neuroscience as part of psychology

what do these examples tell us about the relation between mental phenomena and physical states of the brain? are they:

if we would reach a complete understanding of all brain mechanisms, from the single molecule to the function of extended neural networks, would we then be able to comprehensively predict mental states? neuroscientist A. Snyder (Centre for the Mind) says: Yes --- strong version of neurone doctrine

Summary: current approaches in neuroscience

key reading:

full refernce list:

some study questions

last update 16-04-2005
Johannes M. Zanker