template

advertisement
The pyramid approach to research in
cognitive neuroscience
Grega Repovš
Department of psychology, University of Ljubljana
Aškerčeva 2, SI-1000 Ljubljana
e-mail: gregor.repovs@uni-lj.si
ABSTRACT
The development of a comprehensive model of human
cognition is possible only through cooperation and
exchange of information within the disciplines of
cognitive neuroscience. Each discipline provides a
different but complementary view of the subject of study,
research methods and techniques. Building on
frameworks developed by Marr (1982), Flanagan (1992)
and Kosslyn (1996), we propose a pyramid approach to
research in cognitive neuroscience that tries to combine
the advantages and contributions of constituent fields –
cognitive
psychology,
computational
modelling,
neuroscience and cognitive neuropsychology - into a
powerful research strategy.
Human mind is one of the most exciting and complex
subjects of scientific study. Its study probably spans back to
the beginnings of history, to the emergence of selfawareness. It has been and still is a subject of study of
various disciplines from psychology to physics, from
philosophy to economics. The way people perceive
themselves and the world around them, think about it, reason
about it, decide and solve problems, shapes every moment of
their personal lives, influences the dynamics of social groups,
the success of business, the development of sciences, the
well-being of countries and the future of planet Earth. Each
discipline approaches the subject of human mind and brain
from its own specific point of view, using the methods,
conceptual tools and research paradigms developed through
the history of the discipline and adjusted to the specific
questions it tries to find answers to. Even though addressing
the same subject, the theories, models and findings of
different disciplines were and still are hard if sometimes not
impossible to relate to one another. In the past not many even
considered it worthwhile. It was with the rise of the cognitive
paradigm that things started to change.
The cognitive paradigm brought about two important
changes. First, after the long and hard rule of behaviorism
that flatly denied any validity to the study of human mind,
the mind and its relation to the brain became the central
subject of study, bringing together many disciplines that used
to be either labeled as unscientific or previously didn’t
explicitly address the subject. A broad movement under the
name Cognitive science was born. The second change was a
lot more subtle but arguably even more important. The
cognitive paradigm has established decompositional analysis
as the basic research and explanatory strategy, and
information-processing as its core approach (Atkinson,
1998). Information-processing approach tries to explain how
and why humans possess their capacities and properties in
terms of characteristics of parts of their cognitive system. It
assumes that the brain enables a person to have these
capacities and properties in virtue of having specific
informationprocessing components operating in a specific
way. It was this change that enabled different disciplines to
adopt a common research paradigm, to use the same theories
and models and relate their research findings. Cognitive
psychologists were able to build computational models and
computer simulations of their theories. Neuropsychologists
could use models of normal brain function to explain
cognitive dysfunctions after brain damage and plan
appropriate rehabilitation programs. Neuroscientists were
able to use models of functional architecture to guide their
research efforts in finding relevant neuroanatomical
structures and neurophysiological mechanisms, and at the
same time provide cognitive psychologists with important
ideas and constraints for their theories. Cognitive
neuroscience was born.
Even though possible, the exchange of information and
combination of research approaches within fields of cognitive
neuroscience was not that common or even readily accepted.
Some for instance have insisted that the task of cognitive
psychology is to propose and test models of functional
architecture and not to speculate on structural and material
underpinnings. Even though combination and confrontation
of evidence from different fields of cognitive neuroscience
enabled some important breakthroughs, no generally
accepted framework that would specify how to combine the
information existed. The need for a research framework that
would propose a way to coordinate research efforts and relate
the findings from different fields of cognitive neuroscience in
development of complex theories of human brain function
soon became apparent.
A descriptive proposal and an argument for such a
framework in the study of consciousness was put forward by
Owen Flanagan (1992), and called “the natural method”.
Flanagan argued that to be able to provide a promising theory
of consciousness one has to treat three lines of analysis,
pertaining to three different levels of description, with equal
respect. First, one has to take into account phenomenology
and listen carefully to what individuals have to say about
their experiences. Second, one has to turn to psychologists
and cognitive scientist and their accounts of how mental life
works and what role consciousness has in its overall
economy. Finally, one has to listen to the neuroscientists’
explanations of how different sorts of conscious mental
events are realized in the brain. The object of the natural
method is then to see if the three stories can be rendered
coherent, meshed and brought into relative equilibrium
without any prejudice or a priori decision about which line of
analysis is the best or the most correct one. Flanagan
argued that none of the approaches can “make it” on their
own. Phenomenology can provide detailed descriptions of the
quality of experiences, but can shed no light on the mental
events and processes involved in producing them.
Explanations at the psychological level can provide
illuminating models of mental activity, but as there always
exist a number of different and conflicting functional
accounts of the facts, they need to be constrained by
knowledge of the brain. Neuroscience can provide those
constrains, but the study of the brain alone cannot provide
any knowledge about the mind until psychologists and
phenomenologists describe what phenomena there are to be
explained.
The question of how to relate disciplines within
cognitive neuroscience was actually addressed by David
Marr already in 1982. Marr (1982) outlined three levels at
which a successful information-processing theory should be
specified: computational theory, algorithmic level and
hardware level. At the level of computational theory one
specifies what needs to be computed in order to complete a
specific task; what inputs need to be transformed to what
outputs. The theory at algorithmic level should describe the
exact way the computation is carried out. The representations
used and algorithms that operate on these representations
should be specified in detail. While the computational theory
characterizes the problem to be solved, the theory of
algorithm specifies a possible solution to the problem. At the
lowest, hardware level one specifies how the algorithm is
actually implemented in the brain.
The framework proposed by Marr (1982) presented an
important step forward, but nevertheless received a fair
amount of criticism. Kosslyn (1996) for instance recognizes
the enormous advance and utility of the proposed framework,
but also identifies several difficulties. His first objection is
that the distinction between theory of computation and theory
of algorithm is not always clear. The second objection
pertains to Marr’s separation between computation and it’s
implementation. Whereas Marr implies that one can work at
the computational level without concern for actual
implementation, Kosslyn points out that mental processes are
nothing more - or less - than descriptions of specific brain
function. To Kosslyn it seems odd to formulate a theory of
and neuroanatomical facts.
To overcome the difficulties associated with Marr’s
(1982) hierarchy, Kosslyn (1996) proposes to reconstruct his
approach in the form of a triangle (figure 1). The top of the
triangle represent abilities, by which he understands things
the organism can do, the capacities and properties that are to
be explained. The other two vertices represent a particular
explanation of the abilities offered by two kinds of
considerations, one based on computation, the other on
physiology and anatomy of the brain, with the goal to not
only understand how a set of computation can produce an
observed ability, but to also understand how the brain
performs such computations. As Kosslyn points out, each
consideration interacts directly with the other two, providing
feedback, supplementing and constraining them. The theory
is then developed starting coarse from all ends and
converging on a precise characterization.
In fall 1999, a small informal group of teaching
assistants and students at the Department of Psychology,
University of Ljubljana, started to meet on a regular basis.
Brought together by a common interest in cognitive
neuroscience our goal was to form an efficient study and
research group. The first step that we took was to get
acquainted with existing paradigms and research approaches
in order to be able to formulate common research, goals and
framework that would enable us to relate quite varied
personal research interests and approaches. The Kosslyn’s
(1996) cognitive neuroscience triangle seemed to be just
what we were looking for and we happily adopted it. But as
we were discussing different phenomena, theories and
research findings, we started to feel that something was
missing.
One of the fields where the ability to relate findings from
different fields and research approaches was most fruitfully
used is cognitive neuropsychology. The goal of cognitive
neuropsychology is to use existing theories of normal human
cognition in explaining cognitive dysfunctions following
brain damage. At the same time it provides crucial data for
testing those same theories, as well as new ideas and
constrains for their development. Cognitive neuropsychology
together with functional brain imaging has been at the
forefront of relating functional architecture of the mind and
the structure of the brain and helped in resolving important
theoretical problems, such as the basic structure of human
memory.
Even though it provides important information that can
help guide empirical and theoretical research in all of the
three basic fields of cognitive neuroscience, cognitive
Figure 1. The cognitive neuroscience triangle (adopted from Kosslyn, 1996).
neuropsychology as the study of dysfunctional brain is not
explicitly included in Kosslyn’s cognitive neuroscience
triangle. As we believe that every complete model of
cognitive function should be able to account for cognitive
dysfunctions of the damaged brain, be it in a form of a
functional architecture, computational or structural model,
we have decided to include it as a fourth founding member of
a comprehensive cognitive neuroscience research framework.
As it does not present a new level of description and it relates
to each of the other fields, the logical decision was to include
it not as a fourth point of a rectangle but orthogonaly to the
existing triangle, thus forming a triangular cognitive
neuroscience pyramid. To reflect the fact that each corner of
the pyramid is covered by a different field of cognitive
neuroscience,
characterized
by
different
research
methodologies and somewhat different assumptions and
goals, and at the same time to emphasize the interdependence
and the need for cooperation and multidisciplinary research
in cognitive neuroscience, we have also replaced the subjects
of study or considerations with the names of the fields
studying them (figure 2).
h a v e to study the cognitive dysfunctions following
a specific brain damage to provide further constraints and
tests of the theory, which is the task of cognitive
neuropsychology. Each of the stated lines of inquiry is
related to every other one. There is no proper or ideal
sequence of undertaking the proposed fields of research since
new findings in any of these fields will constantly provide
new tests, constraints and ideas for the rest of them. The
ultimate goal of this research endavour, as Kosslyn (1996)
describes it, is to understand the workings and relations of the
functional and structural components of the brain so well as
to be able to put forward a mechanical model of brain
function and write a computer program that would simulate it
in detail. He believes that the goal will be unreachable for
some time to come. For the time being we should satisfy
ourselves with the broad understanding of the components,
the processes and their interactions, which would enable us to
build a coarse computer simulation. To demonstrate the way
in which the cognitive neuroscience pyramid can be
employed and how findings from different lines of research
can be related and combined in a constructive way, we have
presented examples from the fields of motion perception
(Poljanšek, 2001), working memory (Fesel Martinčevič &
Ravnik, 2001), semantic memory (Repovš, 2001) and
appraisal processes in emotion (Černetič, 2001). We have
also included an overview of use of reaction times across the
cognitive neuroscience pyramid (Bajec & Zakrajšek, 2001) to
indicate how research paradigms can be implemented in
different fields and provide a bridge between them.
References
Figure 2. The fields of research constituting the pyramid
approach to research in cognitive neuroscience.
Let us briefly summarize the main assumptions, goals
and starting points of the cognitive neuroscience pyramid
(i.e. the pyramid approach to research in cognitive
neuroscience) as we propose it. The pyramid approach stands
on the assumption that four lines of scientific inquiry are
necessary for developing comprehensive models of human
mind/brain. First, we have to describe in detail what
capacities and properties the model is to explain and put
forward a functional architecture in virtue of which the
mind/brain enables the person to have those properties and
capacities. This is the role of cognitive psychology. Second,
we have to build working computational models of cognitive
functions that specify in detail the proposed theoretical
assumptions. That is the domain of computational science.
Third, we
have to
t a k e
i n t o
a c c o u n t
t h e
a c t u a l
a n a t o m i c a l
s t r u c t u r e
a n d
p h y s i o l o g y
o f
t h e
b r a i n
t o
b e
a b l e
t o
e x p l a i n
h o w
t h e
c o g n i t i v e
f u n c t i o n s
a r e
a c t u a l l y
i n s t a n t i a t e d
i n
t h e
b r a i n .
N e u r o s c i e n c e
i s
s u p p o s e d
t o
d o
t h a t .
F o u r t h ,
w e
Atkinson, A. P. (1998). Wholes and their parts in cognitive
psychology: Systems, subsystems, and persons.
http://www.soc.unitn.it/dsrs/IMC/IMC.htm.
Bajec, B. and Zakrajšek, K. (2001). Some examples of using
reactiontimes experiments in cognitive neuroscience
pyramid approach. In D. Vodušek, & G. Repovš (Eds.),
Cognitive neuroscience. Information Science 2001,
multiconference proceedings. Ljubljana.
Černetič, M. (2001). Investigating the role of appraisal
processes in emotion: benefits from the pyramid
approach to cognitive neuroscience. In D. Vodušek, &
G. Repovš (Eds.), Cognitive neuroscience. Information
Science 2001, multiconference proceedings. Ljubljana.
Fesel-Martinčevič, M. and Ravnik, J. (2001) Cognitive
neuroscience pryramid to study of working mamory. In
D. Vodušek, & G. Repovš (Eds.), Cognitive
neuroscience.
Information
Science
2001,
multiconference proceedings. Ljubljana.
Flanagan, O. (1992). Consciousness reconsidered.
Cambridge: MIT Press.
Kosslyn, S. H. (1996). Image and brain: The resolution of the
imagery debate. Cambridge: MIT Press.
Marr, D. (1982). Vision: A computational investigation into
the human representation and processing of visual
information. San Francisco: W. H. Freeman.
Poljanšek, A. (2001). Integrative approach to investigation in
motion perception. In D. Vodušek, & G. Repovš (Eds.),
Cognitive neuroscience. Information Science 2001,
multiconference proceedings. Ljubljana.
Repovš, G. (2001). Mechanisms and structure of semantic
memory. In D. Vodušek, & G. Repovš (Eds.), Cognitive
neuroscience.
Information
Science
2001,
multiconference proceedings. Ljubljana.
Download