What Is Cognitive Neuroscience?
Cognitive NeuroscienceThe Mind-Brain ProblemWhat is cognitive neuroscience?Neuroscience is a physical science -- it seeks tounderstand physical mechanisms of the nervoussystem.Cognitive neuroscience is the branch of neurosciencethat seeks to understand the mechanisms of thenervous system that are directly related to cognitive(mental) processes. These mechanisms are thought toreside in the brain. Because cognition refers tofunctions of the mind, we must begin our study ofcognitive neuroscience by first examining the relationbetween the mind and the brain.The question of how the mind and brain are related iscalled the mind-brain problem. It taps into some deepphilosophical issues.
OntologyOntology is the philosophical study of the nature ofreality. It addresses the question of "what exists". Abasic question of ontology is whether there is morethan one order (domain, realm) of reality.Most people, and all scientists, agree that there is aphysical order of reality. It includes all entities andeffects that are described by physical science. Thebrains of humans and other animals are such entities.Within the brain is a physical (neural) order.Controversy arises in deciding whether the physical isthe only realm of existence.Is the mental order separate from, and independent of,the physical order?What makes the mind-brain problem difficult is ntial descriptions are usually called "mental",but it is difficult to determine whether mental, or morecurrently "cognitive", descriptions refer to a unique andseparate domain of existence, or simply are aroundabout way of referring to the physical domain.
The philosophy of mind is a branch of philosophy thatstudies the nature of the mind, and its relationship tothe physical body, particularly the brain. The mindbrain problem (formerly, the mind-body problem) is acentral topic in the philosophy of mind.Although many well-developed philosophies of mindhave been proposed over the centuries, there is still nogenerally agreed-upon solution to the mind-brainproblem.
Philosophy of mind3 traditional approaches to the mind-brain problem:1) Dualism: physical and mental are twofundamental domains of existence.The three main types of dualism differ in the causalrelations they propose between physical andmental phenomena.a. Interactionism: there are physical effectscaused by the mental realm and mentaleffects that have physical causes.b. Epiphenomenalism: causation only occurs inone direction, i.e. from physical to mental.c. Parallelism: mental and physical effects arerelated, but not causally. Mental andphysical events are in directcorrespondence, but do not cause oneanother.2) Idealism: the fundamental domain of reality is themental -- the physical world is the construction ofthe mind -- material objects have no existenceexcept as the contents of perceptual states of themind.3) Physicalism (materialism): the fundamentaldomain of reality is the physical. Mental events
are essentially physical in nature. Different typesof physicalism take different positions regardingthe physical basis of the mind.a. Nonreductive physicalism: mental states arephysical states, but mental states cannot bereduced to behavior, brain states, orfunctional states. [Mental states are physicalbut cannot be reduced to any other physicalstates.]b. Identity theory of mind: mental events areidentical to physical events in the brain.[Mental states can be reduced to brainstates.]c. Eliminative materialism: all mental events willeventually be described as physical eventsin the brain, but only with the elimination of“common-sense” descriptions of mentalphenomena. [An extreme form of the identitytheory saying that mental states eventuallywill all be eliminated by being reduced tobrain states.]d. Behaviorism: the “mind” is a hypotheticalconstruct; mental events are descriptions ofbehavior, not “interior states”. [The “mind” isan unnecessary descriptive term.]
e. Functionalism: the mind is “what the braindoes”; mental events are characterizations ofphysical states of the brain, describing theircausal relation with (functional role in) othermental states, sensory inputs, andbehavioral outputs. By emphasizing function,this theory considers the possibility thatcognition may be manifested by physicaldevices, including computers. Functionalismincludes the computational theory of mind.[Mind function is brain function; mind & brainare related by having complementaryfunctions.]f. Embodied Cognition: the mind is shaped bythe body outside the brain. [Mind functiondepends on body function, including brainfunction.]
Computational mindA functionalist approach to the mind-brain problemhas been proposed by Jackendoff (1990), whodistinguishes between:(a)the phenomenological notion of mind, whichpertains to the mind as the seat of consciousawareness.and(b)the computational notion of mind, whichtreats the mind as an information-bearing andinformation-processing system.From this perspective, we can say that cognitiveneuroscience studies the computational mind-brainrelation.It regards the computational mind as an abstractspecification of functional organization in the nervoussystem.
A matter of correlationEven if we take a functionalist approach and reducethe mind-brain problem to the computational mindbrain problem, then we are still left with the problemthat theories of the computational mind in cognitivescience and theories of the brain in neurosciencerepresent two independent systems of description.Cognitive neuroscience has not developed to thepoint where it has established causal relationsbetween cognitive phenomena and neuralphenomena.All science undergoes a natural progression fromobservation to correlation to causation. Cognitiveneuroscience is largely still at the stage of correlation.Even so, correlation is not a simple matter. It isdifficult to know which neural entities correlate withwhich cognitive entities.
The cognitive neuroscience triangle (Kosslyn& Koenig 1992)To approach this problem, cognitive neuroscienceattempts to establish correlations between cognitivephenomena and neural phenomena, using 3 majordomains:(a) cognition (behavior & models)(b) brain (neurophysiology & neuroanatomy)(c) computation (analyses & models)
Emergence of the Distributed Paradigm inNeuropsychologyTo understand the difference between the modular andnetwork paradigms, it is necessary to examine thehistory of understanding the relation between brainfunction and cognition. This history has beendominated by two parallel trends: localizationism andglobalism, and has led to the emergence of acompromise paradigm, the distributed paradigm.
Localizationism vs globalismHistorically, there has been a controversy for about200 years in neuropsychology over the question ofwhether different mental functions are carried out bydifferent parts of the brain (localizationism) or the brainworks as a single, integrated whole (globalism).PhrenologyIn the 17th & 18th centuries, the theory of faculties wasdominant in psychology.
All psychological processes were understood as"faculties" of mind, incapable of further subdivision.In 1796, Franz Joseph Gall began measuring bumpson the heads of Viennese residents. He postulated thatthe brain is a collection of centers corresponding tospecific "faculties".He thought that even very elaborate & abstractfunctions e.g. cautiousness, generosity, hope, werediscretely localized to single areas of cerebral cortex.Cranial bumps were thought to reflect development ofcortical area underneath and consequently thecorresponding mental trait.This concept, later called phrenology by Spurzheim,represents an extreme expression of the localizationistview.Incorrect assumptions:a) cognitive functions are implemented by discretecortical regionsb) development of a cognitive function increases thesize of its regionc) enlargement of cortical regions causes expansion ofthe outer cranial surface
Correct assumptions:a) mental abilities can be specified and analyzedb) the cerebral cortex is important for mental abilityc) the brain is not a single, undifferentiated system
GlobalismPhrenology was criticized by Pierre Flourens (1824)who found that mental functions are not localized, butthat the brain acts as a whole for each function.The Paris Academy of Sciences commissioned him toinvestigate the claim of Gall that character traits arelocalized in specific cortical regions.He studied the effects of brain lesions on the behaviorof pigeons.The pigeons could recover after parts of the brain wereremoved, regardless of the location of the damage.He concluded that the major brain divisions areresponsible for different functions.Cerebral cortex: perception, motricity, judgmentCerebellum: equilibrium, motor coordinationMedulla: respiration, circulationHowever, he found no localization of cognitive functionwithin the cerebral cortex. He concluded that the cortexhas equipotentiality for cognitive function: lost functionwith ablation does not depend on the location ofdamage, but only on the amount of tissue lost.
The controversy continuedLater animal studies showed that different parts of thebrain do have specific functions:In 1870, Eduard Hitzig (assisted by Gustav Fritsch)supported the localizationist view based on evokedmuscular responses from direct stimulation of thefrontal lobe of the cortex of a dog.In 1881, Hermann Munk removed parts of the occipitallobe of a dog's brain & found that it could still see butcould no longer recognize objects.Clinical evidence suggesting localization of functionalso appeared:In 1861, Paul Broca showed that a lesion of theposterior third of the left inferior frontal gyrus causes ng of speech. He believed that the "motorimages of words" are localized in this part of the brain.In 1874, Carl Wernicke described a patient who haddifficulty comprehending speech after damage to theleft superior temporal gyrus.
Friedrich Goltz - 1881 – was the major opponent oflocalizationism of his time; he postulated that brainworks as a whole. He claimed that Hitzig’s results werebehaviorally irrelevant since the total paralysis onewould expect from ablation of a real motor controlcenter never occurred.During the 1st half of 20th century, several influentialneuroscientists continued to advocate globalism. KarlLashley was most important. He proposed twoprinciples of brain function:a) mass action: the brain works as a single systemb) equipotentiality: all parts have equal ability toperform different tasksHe based his ideas on a long series of experiments totry to find the locus of learning by studying mazelearning in rats with various brain lesions. He declaredthat brain function is widely distributed because hecouldn't find such a locus. He concluded that only theextent of damage was important, not the location.However, maze learning involves many complex motor& sensory capabilities. Even when deprived of 1capability, animals learn with another.Other studies, notably those using electrical stimulationof exposed cortex in awake patients by Wilder Penfield& colleagues, continued to provide evidence oflocalization.
Seeds of resolution: The distributed viewThree individuals (a neurologist, a neuropsychiatrist,and a psychologist) are important for contributing to thedistributed view of brain function that is important forhaving led to the modern network paradigm.1) John Hughlings Jackson was an English neurologistwho contributed to neurology and psychology from1861 to 1909. He developed a theory of evolutionaryneuropsychology, in which 3 evolutionary levels arefound in the nervous system. Functions are distributedat each level, and across the 3 levels.Jackson proposed that cortical lesions cause“negative” symptoms (as well as positive ones), inwhich the loss of cortical control releases the samefunction at a lower level. He gave the example of abrain-injured patient who could not voluntarily speak,but could emit speech involuntarily, i.e. even though hecould not find the word for a simple object, he couldstill swear vigorously when provoked. Thus, Jacksonargued against a strict localizationist view of brainfunction.
2) Wernicke (1874) also proposed the idea thatcomplex functions (e.g., language) are composed oflocalizable simple perceptual and motor functions:a) complex functions are composed of separatecomponents -- proposed that language is not a singlefunction, but has at least 2 components (i.e.comprehension and articulation).b) functions localized in distinct brain areas are notcomplex attributes as postulated by phrenologists, butare much simpler perceptual and motor functions.3) The idea of a distinction between complex andelementary functions was supported by Lev Vygotskii(1934), a Russian psychologist who emphasized thedevelopmental nature of complex psychologicalfunctions. They are not elementary and indivisible, butrather they may change their composition from onestage of development to the next.
Summary of distributed viewThe distributed view was clearly articulated byAlexander Luria (1975): "The higher forms of humanpsychological activity and all human behavioral actstake place with participation of all parts and levels ofthe brain, each of which makes its own specificcontribution to the work of the functional system as awhole."1) Elementary functions are localized, but the brainworks in a distributed manner to produce complexfunctions that are not localized.Why are elementary functions localized to particularbrain areas?Because neurons that perform an elementary function:a) receive from the same input sourcesb) project to the same output targetsc) must interact quickly2) Complex functions are carried out by distributedcombinations of simple functions. The simple functionsare localized in many different places in the brain. Theycan be carried out by different elementary functions atdifferent times, allowing them to be performed indifferent ways. Thus, different "strategies" can beimplemented as different combinations of simplefunctions.
Resolving elements of localizationism andglobalism, the distributed view has evolved intothe modern network paradigm inneuropsychology.
History of Neural Networks in ArtificialIntelligenceThe concept of “neural network” in artificialintelligenceTo understand the computational aspects of thenetwork paradigm requires examining the history of theconcept of “neural network” in the field of artificialintelligence.The modern history of artificial intelligence can betraced back to the 1940's, when 2 complementaryapproaches to the field originated.The Serial Symbol Processing (SSP) approach beganin the 1940's, when the architecture of the moderndigital computer was designed by John von Neumannand others. They were heavily influenced by the workof Alan Turing on finite computing machines. TheTuring Machine is a list of instructions for carrying out alogical operation.The von Neumann computer follows this theme. It:a) performs one operation at a timeb) operates by an explicit set of instructionsc) distinguishes explicitly between stored information &the operations that manipulate information.
The Parallel Distributed Processing (PDP) approach(also called connectionism) may also be traced to the1940’s.In 1943, Warren McCulloch and Walter Pitts proposeda simple model of the neuron – the linear thresholdunit. The model neuron computes a weighted sum ofits inputs from other units, and outputs a one or zeroaccording to whether this sum is above or below athreshold.McCulloch & Pitts proved that an assembly of suchneurons is capable in principle of universalcomputation, if the weights are chosen suitably. Thismeans that such an assembly could in principleperform any computation that an ordinary digitalcomputer can.
In 1949, Donald Hebb constructed a theoreticalframework for the representation of short-term &long-term memory in nervous system.The functional unit in Hebb's theory is the NeuronalAssembly: a population of mutually excitatory neuronsthat when excited together becomes functionallylinked.He also introduced the Hebbian learning rule: whenunit A and unit B are simultaneously excited, thestrength of the connection between them is increased.A leading proponent of the PDP approach was FrankRosenblatt.In the late 1950’s, he developed the concept of theperceptron: a single-layer network of linear thresholdunits without feedback.The work focused on the problem of determiningappropriate weights for particular computational tasks.For the single-layer perceptron, Rosenblatt developed
a learning algorithm – a method for changing theweights iteratively so that a desired computation wasperformed. (Remember that McCulloch & Pitts hadproposed that the weights in their logic circuits had tobe appropriate for the computation.)The properties of perceptrons were carefully analyzedby Minsky & Papert in their 1969 book "Perceptrons".They showed that Rosenblatt’s single-layer perceptroncould not perform some elementary computations. Thesimplest example was the “exclusive or” problem (theoutput unit turns on if 1 or the other of 2 input lines ison, but not when neither or both are on).
Rosenblatt believed that multi-layer structures couldovercome the limitations of the simple perceptrons, buthe never discovered a learning algorithm fordetermining the way to arrive at the weights necessaryto implement a given calculation.Minsky & Papert’s analysis of the limitations of onelayer networks suggested to many in the fields ofartificial intelligence and cognitive psychology thatperceptron-like computational devices were not useful.This put a damper on the PDP approach, and the late1960's and most of the 1970's were dominated by theSSP approach & the von Neumann computer.
During this time, many grandiose claims for the SSPapproach were not fulfilled. Also, the backwardpropagation of error technique was discovered.These developments led to a resurgence of interest inPDP models in the late 1970's.It was realized that, although Minsky & Papert wereexactly correct in their analysis of the one-layerperceptron, their analysis did not extend to multi-layernetworks or to systems with feedback loops.The PDP approach has gained a wide following sincethe early 1980's.Many neuroscientists believe that it embodiesprinciples that are more neurally realistic than the SSPapproach. Because PDP models are thought to worklike brain regions, they are often called artificial neuralnetworks.
Properties of artificial neural networks1) Artificial neural networks (ANNs) are organized aslayers of units.2) A feedforward network has an input layer, an outputlayer, and one or more hidden layers.3) Each unit has an output, which is its activity level,and a threshold, which is a level that must beexceeded by the sum of its inputs for the unit to give anoutput.4) Connections between units can be excitatory orinhibitory. Each connection has a weight, whichmeasures the strength of the influence of 1 unit onanother.5) Neural networks are trained by teaching them toproduce certain output when given certain input.Example: training by backward error propagation:(1) randomize the weights(2) present an input pattern(3) compare the output with the desired output (i.e.compute the error)(4) slightly adjust the weights to reduce the error(5) repeat (2) - (4)6) The trained network functions as an associativememory: it relates patterns from the input space tocorresponding patterns in the output space.7) The network can also be considered to perform amapping of input space to output space.
8) The pattern of weights on the internal connections ofthe network can be considered to be a representation:they represent the combinations of input features thatidentify output patterns.9) A recurrent network has excitatory or inhibitoryfeedback connections from higher units back to loweruni
Cognitive neuroscience is the branch of neuroscience that seeks to understand the mechanisms of the nervous system that are directly related to cognitive (mental) processes. These mechanisms are thought to reside in the brain. Because cognition refers to functions of the mind, we must begin our study of
Cognitive neuroscience is the branch of neuroscience that seeks to understand the mechanisms of the nervous system that are directly related to cognitive (mental) processes. These mechanisms are thought to reside in the brain. Because cognition refers to functions of the mind, we must begin our study of cognitive neuroscience by first examining .
Cognitive Neuroscience Philipp Koehn 11 February 2020 Philipp Koehn Artificial Intelligence: Cognitive Neuroscience 11 February 2020. Cognitive Neuroscience 1 Looking ”under the hood” What is the hardware that the mind runs on? Much progress in recent years
theory of mind, empathy, emotion regulation, self-control, mirror neurons, social cognition, social neuroscience, automaticity, neuroeconomics Abstract Social cognitive neuroscience examines social phenomena and pro-cesses using cognitive neuroscience research tools such as neu-roimaging and neuropsychology. This review examines four broad
social cognitive neuroscience, and many of the attendees have become leaders in the field, despite few having pub-lished social cognitive neuroscience findings at that point. There were introductory talks on social cognition and cog-nitive neuroscience by Neil Macrae and Jonathan Cohen, respectively, along with symposia on stereotyping (William
Social cognitive neuroscience: where are we heading? Sarah-Jayne Blakemore1, Joel Winston2 and Uta Frith1 1Institute of Cognitive Neuroscience, 17 Queen Square, London, WC1N 3AR, UK 2Wellcome Department of Imaging Neuroscience, 12 Queen Square, London, WC1N 3BG, UK Humans crave the company of others and suffer pro-foundly if temporarily isolated from society.
Katie Marie Barnes Ronceverte, WV CLINICAL NEUROSCIENCE Robert Gates Bass III Midlothian, VA EXPERIMENTAL NEUROSCIENCE Meet the Seniors Ben Batman Monroe, VA CLINICAL NEUROSCIENCE Kourtney A Baumfalk Richmond, VA COGNITIVE AND BEHAVIORAL NEUROSCIENCE Jennifer Leigh Beauchamp West Chester, PA CLINICAL NEUROSCIENCE
neuroscience, learning, cognitive neuroscience, and neurorehabilitation. The Behavioral Neuroscience Graduate Program at the University of Alabama at Birmingham (UAB) is one of three Ph.D. granting programs (i.e. Behavioral Neuroscience, Lifespan Developmental Psychology, and Medical/Clinical
Relevant topical seminars offered by Cognitive Neuroscience faculty are also often taught under common/shared course codes for other areas in the department, e.g. Attention & Cognitive Control (7695-Leber), Social Cognitive Neuroscience (7897-Wagner), Human Neuropsychology (8891-Prakash), Neural Dynamics of Decision Making (8890-Krajbich).
