Dalenoort G.J.: THE PARADIGM OF SELF-ORGANIZATION (Gordon &
Breach, 1989)
A collection of articles from experts in various disciplines that
all deal with autonomous systems. Topics include cybernetics,
evolution, complexity, morphogenesis, self-reference. Csanyi
offers a general theory of evolution based on a "replicative
model" of self-organization.
Dalenoort G.J.: THE PARADIGM OF SELF-ORGANIZATION II (Gordon &
Breach, 1994)
The new collection includes articles on learning, the arrow of
time, cellular automata, cognition, etc.
Damasio Antonio: DESCARTES' ERROR (G.P. Putnam's Sons, 1995)
Damasio is trying to build a neurobiology of rationality. In this
book he provides a neurophysiological analysis of memory, emo-
tions and consciousness.
The book has three themes. 1. Human reason depends on the
interaction among several brain systems rather than on a single
brain centre. 2. Feelings are views of the body's internal
organs. Feelings are percepts and they are as cognitive as any
other percept. 3. The mind is about the body: the neural
processes that are experienced as the mind are about the
representation of the body in the brain. The mental requires the
existence of a body for more than mere support: the mind is not
a phenomenon of the brain alone. The mind derives from the entire
organism as a whole. The mind reflects two types of interaction:
between the body and the brain, and between them and the environ-
ment.
Damasio formulates the somatic-marker hypothesis: a special class
of feelings, acquired by experience, express predicted future
outcomes of known situations and help the mind make decisions.
The neural basis for the self resides with the continous reac-
tivation of 1. the individual's past experience (which provides
the individual's sense of identity) and 2. a representation of
the individual's body (which provides the individual's sense of a
whole). The self is continously reconstructed. This is a purely
non-verbal process: language is not a prerequisite for conscious-
ness. Nonetheless, language is the source of the "I", a second
order narrative capacity. Damasio's "embodied mind" is closely
related to Edelman's "self imbued with value".
Damasio's theory of convergence zones (not presented in this
book) is tackling the issue of consciousness. When an image
enters the brain via the visual cortex, it is channelled through
"convergence zones" in the brain until it is identified. Each
convergence zone handles a category of objects (faces, animals,
trees, etc): a convergence zone does not store permanent memories
of words and concepts but helps reconstructing them. Once the
image has been identified, an acoustical pattern corresponding to
the image is constructed by another area of the brain. Finally an
articulatory pattern is constructed so that the word that the
image represents can be spoken. There are about twenty known
categories that the brain uses to organize knowledge:
fruits/vegetables, plants, animals, body parts, colors, numbers,
letters, nouns, verbs, proper names, faces, facial expressions,
emotions, sounds. "Convergence zones" are indexes that draw
information from other areas of the brain. The memory of some-
thing is stored in bits at the back of the brain (near the gate-
ways of the senses): features are recognized and combined and an
index of these features is formed and stored. When the brain
needs to bring back the memory of something, it will follow the
instructions in that index, recover all the features and link
them to other associated categories. As information is pro-
cessed, moving from station to station through the brain, each
station creates new connections reaching back to the earlier lev-
els of processing. These connections always allows the brain to
work in reverse. Convergence zones may be common to all indivi-
duals or different from individual to individual, based on
experience.
Emotions are the brain's interpretation of reactions to changes
in the world. Emotional memories involving fear can never be
erased The prefrontal cortex, amygdala and right cerebral cortex
form a system for reasoning that gives rise to emotions and feel-
ings. The prefrontal cortex and the amygdala process a visual
stimulus by comparing it with previous experience and generate a
response that is transmitted both to the body and to the back of
the brain.
Convergence zones are organized in a hierarchy: lower convergence
zones pass information to higher convergence zones. Lower zones
select relevant details from sensorial information and send sum-
maries to higher zones, which successively refine and integrate
the information. In order to be conscious of something a higher
convergence zone must retrieve from the lower convergence zones
all the sensory fragments that are related to that something.
Therefore, consciousness occurs when the higher convergence zones
fire signals back to lower convergence zones.
Davidson Donald: INQUIRIES INTO TRUTH AND INTERPRETATION
(Clarendon Press, 1984)
Davidson is the main proponent of "truth-conditional semantics",
which asserts the central place in the theory of meaning of a
theory of truth.
With his "anomalous monism", Davidson promotes the token theory
of identity: the same instance of a mental state may correspond
to different neural states at different times. Given a mental
state, it is not possible to relate it to a specific physical
state. The same event may be both mental and physical, but there
is no relationship between the two descriptions. There cannot be
any relationship between the psychological vocabulary and the
neurophysiological vocabulary.
Davidson's theory of the mind rests on three principles. At
least some mental events interact causally with physical events
(causal interaction). Events related as cause and effect fall
under strict deterministic law (the nomological character of
causality). There are no strict deterministic laws under which
mental events can be predicted and explained (the anomalism of
the mind). The physical and the mental realms have essential
features which are somehow mutually incompatible. There can be no
laws connecting the mental with the physical. Therefore there can
be no theory connecting psychology and neurophysiology.
Davidson's conception of the mind is based on the intentional.
Propositional attitudes constitute the basic vocabulary of the
mind. Laws of the mind would then be laws expressed in terms of
intentional expressions.
Davidson thinks that rationality (interpreting agents in terms of
beliefs and desires) provides the sole criterion for psychologi-
cal judgement. His view of the mental is holistic: the attribu-
tion of any mental state to a person requires that the total
system of propositional attitudes be maximally coherent and
rational.
Tarski simply replaced the universal and intuitive notion of
"truth" with an infinite series of rules which define truth in a
language relative to truth in another language. Davidson would
rather assume that the concept of "truth" need not be defined,
that it be known to everybody. Then he can use the corrisponden-
tial theory of truth to define meaning: the meaning of a sentence
is defined as what would be if the sentence were true.
The task for a theory of meaning is then to generate all meta-
sentences (or "T-sentences") for all sentences in the language
through a recursive procedure. This account of meaning only
relies on truth conditions.
A sentence is meaningful in virtue of being true under certain
conditions and not others. To know the meaning of a sentence is
to know the conditions under which the sentence would be true. A
theory of a language must be able to assign a meaning to every
possible sentence of the language. Just like Chomsky had to
include a recursive procedure in order to explain speaker's
unlimited ability to recognize sentences of the language, so
Davidson has to include a recursive procedure in order to explain
speaker's unlimited ability to understand sentences of the
languages.
Natural languages exhibit an additional difficulty over formal
languages: they contain deictic elements (demonstratives, per-
sonal pronouns, tenses) which cause truth value to fluctuate in
time and speaker. Davidson therefore proposes to employ a pair of
arguments for his truth predicate, one specifying the speaker and
one specifying the point in time.
Language transmits information. The speaker and the listener
share a fundamental principle to make such transmission as effi-
cient as possible. Such "principle of charity" asserts that the
interpretation to be chosen is the one in which the speaker is
saying the highest number of true statements. During the conver-
sation the listener tries to build an interpretation in which
each sentence of the speaker is coupled with a truth-equivalent
sentence.
Davies Paul: GOD AND THE NEW PHYSICS (Penguin, 1982)
The book surveys the mysteries of the universe, life, mind, cons-
ciousness, particle physics by updating the debate to the
theories of non-linear dynamics and self-organization.
Davies Paul: ABOUT TIME (Touchstone, 1995)
A popular introduction to relativistic and quantum time, roaming
from big bang to black holes, speculating on time reversal and
tachyons.
Davies Paul: THE MIND OF GOD (Touchstone, 1993)
Davies reviews quantum cosmological theories of the universe and
recent mathematical advances to prove that there is still room
for a God. A wealth of philosophical and scientific notions are
mixed, related and compared. Davies investigates whether the
universe can create itself, the relationship between the world of
Mathematics and the physical world, Artificial Intelligence, etc.
Davis Ernest: REPRESENTATION OF COMMON-SENSE KNOWLEDGE (Morgan
Kaufman, 1990)
A comprehensive and well-organized survey of research areas
related to common sense.
Common sense is a key factor in acting in the real world. Common
sense encompasses both reasoning methods and knowledge that are
obvious to humans but that are quite distinct from the tools of
classical mathematics.
To prepare adequate logical theories for dealing with common
sense, Davis introduces the notation of first-order logic. Essen-
tial to reproducing the power of ordinary language is the use of
operators on sentences. Operators on sentences that apply only to
a limited class of sentences, commute with the quantifiers and
the boolean operators, are referentially transparent and are
closed under inference, are "extensional operators" (e.g., the
temporal operator). Another class of operators on sentences is
that of modal operators (possible and necessary), which obey
their own set of axioms. The meaning of a modal logic is defined
in terms of possible-world semantics.
Classical logic needs also to be extended with plausible reason-
ing: degrees of belief, default rules, inference in the face of
absence of information, inference about vague quantities, analog-
ical reasoning, induction and so forth. A crucial tool for plau-
sible reasoning is non-monotonic logic, which allows inferences
to be made provisionally and, if necessary, withdrawn at any
time. Next, the domain of inference must be somehow closed, and
this can be done in a number of ways: the closed-world assumption
(all relations relevant to the problem are mentioned in the prob-
lem statement), circumscription (extends the closed-world assump-
tion to non-ground formulas as well, i.e. assumes that as few
objects as possible have a given property), default theory (all
members of a class have all the properties characteristic of the
class if it is not otherwise specified). Uncertainties can be
represented with probability theory.
Common sense domains to be dealt with include: physical quanti-
ties (whose values can be ordered, that can be subdivided in
partially ordered intervals, that can be assigned signs based on
their derivaties, whose relations can be expressed in the form of
transition networks, whose behavior can be expressed in the form
of qualitative differential equations); time (whose operators can
be either introduced in a world of discrete, self-contained
situations and events or as part of a modal logic); space (and
the related concepts of distance, containment, overlapping, boun-
daries); physics (according to both DeKleer's component model and
Forbus' qualitative process theory); propositional attitudes
(specifically the relationship between belief and knowledge);
actions (planning systems); and socializing (speech acts.
Davis does not discuss how common sense is learned and whether
some common sense is innate.
Davis thinks that a first-order logic can be endowed with axioms
that reflect the laws of the physical world. Davis' world is made
of a finite set of solid objects that move in space and do not
overlap. Each object has three properties: a mass distribution,
an elasticity coefficient and a friction coefficient. Davis
defines an onthology which includes terms such as: quantity, dis-
tance, objects, and so forth. The theory suggests that an ade-
quate representation of the physical needs to employ Euclides'
geometry, an ontology of space-temporal properties and a set of
axioms about what is going on in the world.
Davis Randall & Lenat Douglas: KNOWLEDGE-BASED SYSTEMS IN ARTIF-
ICIAL INTELLIGENCE (McGraw-Hill,1982)
The first part is devoted to the system AM, that was built to
study and simulate discovery of heuristics in solving mathemati-
cal problems. The second part describes TEIRESIAS, a system to
acquire and maintain large knowledge bases.
Davis Steven: CONNECTIONISM (Oxford University Press, 1992)
An introduction to the field with emphasis on how higher cogni-
tive tasks can be explained by lower connectionist models.
Davis Steven: PRAGMATICS (Ocford University Press, 1991)
An ambitious collection of seminal papers on speech acts (Grice,
Kripke, Searle), indexicals (Kaplan's logic of demonstratives),
implicature and relevance (Grice's "Logic and conversation", Wil-
son & Sperber's "Inference and implicature"), presupposition
(Lewis, Stalnaker), metaphor (Davidson, Searle, Sperber & Wil-
son).
Robyn Carston advances a proposal to distinguish two kinds of
semantics: a linguistic semantics (a theory of utterance) and a
truth-conditional semantics (a theory of propositions). Linguis-
tics semantics provides the input to pragmatics and the two
together provide the input to truth-conditional semantics.
Kent Bach views linguistic communication as an inferential pro-
cess and presents a theory of speech acts.
John Searle attempts to explain "indirect speech acts" in terms
of his theory of speech acts and metaphor as "speaker's utterance
meaning" (a set of principles allow the hearer to compute the
possible meanings).
Dawkins Richard: THE SELFISH GENE (Oxford Univ Press, 1976)
This is one of the books that introduced new methods of thinking
about life, behavior and evolution. Dawkins argues that the gene
is the fundamental unit of evolution. Genes drive evolution and
genes drive behavior.
Darwin's assumption that natural selection favors those individu-
als best fitted to survive and reproduce can be restated as:
natural selection favors those genes that replicate through many
generations.
The level at which selection occurs is not that of the individual
organism, but that of particular stretches of genetic material.
Organisms are merely the means that genes use to perpetuate
copies of themselves.
The universe is dominated by stable structures. And one particu-
lar stable structure is a molecule that makes copies of itself.
Dawkins proves with a number of examples at all levels that self-
ishness is pervasive in nature.
Dawkins also introduces the concept of "memes", the analogous of
genes for cultural transmission. A meme is an idea that repli-
cates itself from mind to mind, such as a slogan or a refrain or
a proverb. Memes behave in a very similar way to genes.
Dawkins Richard: THE EXTENDED PHENOTYPE (OUP, 1982)
The main claim of the book is that the gene is the unit of
natural selection. Genes are selected by their phenotypic
effects. Such phenotypic effects are not limited to individual
organism, but reach out to an "extended" phenotype, consisting of
the world the organism interacts with. Genes ensure their sur-
vival by means of phenotypic effects on the world.
The organism alone does not have biological relevance. What makes
sense is an open system made of the organism and its neighbors.
For example, a cobweb is still part of the spider. The control of
an organism is never complete inside and null outside: there is
rather a continuum of degrees of control, which allows partiality
of control inside (e.g., parasites operate on the nervous system
of their hosts) and an extension of control outside (as in the
cobweb). The genome of a cell can be viewed as a representation
of the environment inside the cell.
Conversely, within the boundaries of an organism there can be
more than one psychology (as in the case of schizophrenics).
The same arguments apply to memes, which are nonbiological repli-
cators. The extended phenotype of a meme is defined by phenotypic
effects such as words, music, images, gestures, fashion, ...
Throughout the book, Dawkins downplays the importance of single
organisms and emphasizes the "extended phenotype" which extends
as far as its control reaches out.
The book is mainly written for biologists and debates numerous
alternative theories.
Dawkins Richard: THE BLIND WATCHMAKER (Norton, 1987)
A very accessible introduction to a variety of topics in evolu-
tionary biology.
The theme of the book is paradox of natural selection, which on
one hand proceeds in a blind and purpose-less way and on the
other hand produces the illusion of more and more complex design.
Dawkins compares biological systems and artificial systems: the
theory of radar vision and the theory of bats' ecolocation
developed in parallel, unaware one of the results of the other,
and eventually formulated the same computational model.
Complex organisms came to be by gradual, cumulative transforma-
tions from simple beginnings. Dawkins emphasizes that darwinism
is not a theory of random chance. Order is created by the "cumu-
lative" property of selection.
Dawkins speculates on the processes that originated life,
reiterates his view that genes are selected by virtue of their
interaction with the environment (including other genes), proves
that punctuated equilibrium is consistent with darwinism, and
compares differing darwinist theories.
Dawkins Richard: RIVER OUT OF EDEN (Basic, 1995)
This is an introduction for the general audience to Dawkins'
ideas and to modern evolutionary theories.
Within his own theory of the genes' struggle and competition for
survival, Dawkins tries to answer philosophical questions such as
how life began and why are we alive at all. Nature's excesses
and cruelties are explained by the need of genes to survive and
reproduce. Suffering, pain and fear to the most horrible
extremes, are part of this game.
DeDuve Christian: VITAL DUST (Basic, 1995)
A detailed and fascinating history of life on the earth, how it
"emerged" and how it developed, from the first catalysts of life
all the way down to the mind.
Delahaye JeanPaul: FORMAL METHODS IN ARTIFICIAL INTELLIGENCE
(Halsted, 1987)
An introduction to recursive functions, Church's thesis, Lambda
calculus, first-order predicate calculus, resolution, unifica-
tion; all the logical tools needed to understand the Prolog pro-
gramming language.
Depew David & Weber Bruce: DARWINISM EVOLVING (MIT Press, 1994)
A competent, comprehensive and exhaustive history and survey of
evolutionary theories from Darwin to Gould and Lewontin.
The first half of this book is a history of darwinism. The
second part deals with Galton, Mendel, Fisher, Wright up to the
modern day synthesis. The third part starts with the discovery of
the DNA and ends with modern models of evolution.
The book shows that the idea of natural selection has undergone
three stages of development, parallel to developments in the phy-
sical sciences: the deterministic dynamics of Isaac Newton, the
stochastic dynamics of Clerk Maxwell and Ludwig Boltzmann, and
now the dynamics of complex systems. If initially Darwin's theory
could be related to Newton's physics in that it assumed an exter-
nal force (natural selection) causing change in living organisms
(just like Newton posited an external force, gravity, causing
change in the motion of astronomical objects), with the invention
of population genetics by Ronald Fisher and others darwinism
became stochastic (the thermodynamic model of genetic natural
selection, in which fitness is maximized like entropy), just what
physics had become with Boltzmann's theory of gases.
Population genetics showed that Darwin's theory (that change
occurred by the natural selection of many minute variations) and
Mendel's theory (that change occurred suddenly, by mutation) were
complementary: changes occur in the frequencies of genes.
The authors point to the dynamics of complex systems, and specif-
ically to the idea of self-organization, as the next step in the
study of evolution.
DeMey Marc: THE COGNITIVE PARADIGM (Univ of Chicago Press, 1982)
Philosophical reflections on the emergence of a new scientific
revolution, the cognitive paradigm.
Dennett Daniel: CONTENT AND CONSCIOUSNESS (Routledge, 1969)
The distinction between mental and physical is ambigous. The dis-
tinction between psychological language and scientific language,
on the other hand, corresponds to the distinction between inten-
tional sentences and extensional sentences. In order to reduce
the mind to the body, one must reduce intentionality to the
extensional.
An extensional reduction of intentional sentences is possible
with internal events serving as the conditions of ascription.
There could be a system of internal states whose extensional
description provides also an intentional description. The problem
is whether it make sense to ascribe content to neural states.
Dennett distinguishes between consciousness (conscious of, non-
intentional sense) and awareness (aware that, intentional sense).
Consciousness is then merely awareness of the contents of inter-
nal states.
Knowledge does not divide into independent parts and therefore
cannot be listed and therefore people can't really say what they
know.
Dennett Daniel: THE INTENTIONAL STANCE (MIT Press, 1987)
Dennett's theory of intentionality is based on the folk concepts
of belief, desire, intention and expectation.
In order to explain and predict the behavior of a system one can
employ three strategies: a "physical stance", which infers the
behavior from the physical structure and the laws of Physics; a
"design stance", which infers the behavior from the function for
which it was designed (we know when a clock alarm will go on even
if we don't know the internal structure of the clock); and an
"intentional stance", which infers the behavior from the beliefs
and desires that the system must exhibit to be rational.
The "intentional stance" is the set of beliefs and desires of an
organism that allow an observer to predict its actions. Belief
and desires are not internal states of the mind which cause
behavior, but simply tools which are useful to predict the
behavior. No system is really intentional.
The process that defines how beliefs and desires are shaped, and
how they affect the organism's behavior, has biological roots. If
an organism survived natural selection, the majority of its
beliefs are true and the way the organism employs them is the
most "rational" (beliefs are used to satisfy its desires).
From a biological standpoint, the intentional stance defines the
relationship between an organism and its environment. The organ-
ism continously reflects its environment, as the organization of
its system implicitly contains a representation of the environ-
ment.
Intentional states are not internal states of the system, but
descriptions of the relationship between the system and its
environment. An intentional state is not separate from the oth-
ers, but, holistically, it makes sense only to deal with the
cognitive state of an organism as a whole, and with its relation-
ship as a whole with the environment. The propositional attitude
is defined by a "notional attitude", which is independent of the
real world, and a component which depends from the real world.
A notional attitude is defined in a "notional world". An agent's
notional worlds are the worlds in which all the agent's beliefs
are true and all the agent's desires are feasible. Me and my dop-
pelganger on Putnam's twin Earth have the same notional world,
but different propositional attitudes (because we live in two
different environments).
Intentionality defines an organism as a function of its beliefs
and desires, which are products of natural selection. The more an
agent's notional worlds stride away from the real world, the less
the agent is capable of adapting to it.
What creates beliefs and desires is the biological function of
cognitive mechanisms. Beliefs must be true and desires must be
feasible to be useful to survival.
However, Brentano's thesis (that the intentional is irreducible
to the physical) is true, because strickly speaking there are no
such things as beliefs and desires.
Dennett's theory allows for an interpretation within an ecologi-
cal context, in agreement with Gibson's and Neisser's theories;
within an ethological context (cognitive profile of a species);
and within a philogenetic context (how an organism evolved do
adapt continously to its environment).
Dennett Daniel: CONSCIOUSNESS EXPLAINED (Little & Brown, 1991)
Dennett's ambition is an empirical theory of the mind. By extend-
ing the Cartesian Theatre (the idea that there is a centered
locus in the brain that directs consciousness) with a multiple
draft model (in which all varieties of perceptions and thoughts
are accomplished by parallel, multitrack brain processes), Den-
nett offers an explanation of how the brain represents time,
anchored around the principle that "probing precipitates narra-
tives" (people are not always conscious of what is happening to
them). Consciousness is spread around the brain and in time.
Consciousness is nonlocalized and nonlinear. Despite the
apparent unity and continuity of our experience, consciousness
does not involve the existence of a single central self, but
arises from the interaction with the environment.
Consciousness exists because it helps survive and it evolved from
non-consciousness to reasoning and then to memes. Dennett thinks
that qualia, and conscious states in general, don't exist. Cons-
ciousness is a collection of memes. The brain is a computer that
collects memes.
The mind must be reduced to a set of cognitive functions. Each
function must be reduced to simpler cognitive problems. And so
forth, each time reducing the intelligence needed to solve the
problem, until we reach a level at which problems can be solved
with no more intelligence than the one that can be found in a
machine. At each level the behavior of a system is given by the
interaction of a set of interconnected components ("homunculi").
Each component's behavior is itself defined by a set of intercon-
nected components.
Having relied massively on Artificial Intelligence ideas, Dennett
also takes aim at Searle's chinese room thought experiment and
attacks each of its three premises.
Dennett Daniel: DARWIN'S DANGEROUS IDEA (Simon & Schuster, 1995)
Dennett offers a personal view of Darwin's contribution to Sci-
ence. Darwin's "dangerous idea" is that design can emerge spon-
taneously via an algorithmic process, since evolution by natural
selection can be viewed as an algorithmic process. A mindless and
mechanical (and relatively simple) process is responsible for
creating the complex systems of life. Design is created at each
run of the algorithm and conserved as the starting point for the
next run. Complex design such as exhibited in living systems is
therefore the product of a process of "accumulation of design"
carried out over time.
Dennett emphasizes that what appears as a very intelligent pro-
cess is in reality made of many tiny stupid steps (he proposed a
similar explanation for the intelligence of the human mind).
Dennett defends James Mark Baldwin's effect, originally proposed
in 1896: that species capable of "reinforced learning" evolve
faster. Unlike Lamarck, who thought organisms can pass on to
their offsprings acquired characteristics, Baldwin thought that
organisms can pass on their capacity to acquire certain charac-
teristics.
The actual genomes that have ever existed are obviously just a
tiny percentage of all the genomes that could possibly exist.
Biological possibility can be reduced to a search (the "tree of
life") in the space of all possible genomes (the "design" space).
An organism is more or less biologically possible if the
corresponding genome can be more or less easily accessed from one
of the existing genomes. There may be local and universal con-
straints (biological laws) that limit the possible routes, just
like there are physical laws that limit which objects can exist.
For example, laws of form may constrain the relation between
genotype and phenotype.
Similar considerations apply to human artifacts, from books to
religions, from languages to Dawkins' memes. They are also
indirectly artifacts of the same process that created living
organisms. Therefore one can conceive of a unified design space
that is navigated by both biological and human creativities.
Dennett discusses how life can have created itself. Self-
replicators are too complex to have occurred by coincidence. He
resorts to hypotheses advanced by Cairns-Smith and Eigen.
Dennett emphasizes that the code reader is as important as the
code: there are infinite ways that the instructions contained in
the DNA could be implemented, and the "decoder" determines which
one will actually be chosen. The message is ambigous and it can
be disambiguated only by the specific decoder that was meant to
decode it. From this observation Dennett concludes that the code
and the decoder must have evolved together. In general, Dennett
argues that "any functioning structure carries information about
the environment in which its function works".
Biology is just another form of engineering.
Dennett strenously defends "adaptionism" against Gould's and
Lewontin's critique (a famous 1979 paper) and argues that it must
form the core of evolutionary biology.
The human species is unique in that it relies on cultural
transmission of information, and such process is carried out by
Dawkins' memes, the units of cultural evolution. The mind was
created when the brain was invaded by memes: memes have created
the mind, not the other way around. Consciousness is therefore a
collection of memes that is implemented in the brain as a sort of
software in a machine that evolved in nature. Dennett in practice
denies the existence of truly conscious states. Meaning itself
is an emergent product of the meaningless algorithm that carries
out evolution.
Dennett defends Artificial Intelligence from Penrose's critique
(based on Godel's incompleteness theorem). Artificial
Intelligence actually fits very well in this scenario of algo-
rithmic (physical and mental) evolution.
DeSousa Ronald: THE RATIONALITY OF EMOTION (MIT Press, 1987)
A study on emotions from a biological (rather than psychological)
perspective.
Emotions are not irrational behavior. They play the same role as
perceptions: they contribute to create beliefs and desires. Emo-
tions are perceptions that play a role in beliefs and desires.
Emotions are learned like a language. Their semantics derives
from the paradigm scenarios in terms of which they have been
learned. The intentionality of emotions leads to a classifica-
tion of objects of emotion. Emotions are also defined by their
relation to time (e.g., an event lasting for years cannot count
as a surprise).
Emotions and reason are not antagonists. Reason and emotion are
complementary cognitive skills. DeSousa talks of "axiological
rationality". Emotions control the crucial factor of salience
and can therefore restrict the combinatorial possibilities that
reason has to face (thereby avoiding the frame problem).
Deutsch, J. Anthony: THE STRUCTURAL BASIS OF BEHAVIOR (Univer-
sity of Chicago Press, 1960)
Deutsch's studies on the rat's behavior reached the conclusion
that rats make purely topological maps of their environment. A
map contains a representation of points in the environment and
conncetions between such representations. A point in the
environment is recognized by comparing its sensory representation
with the representations in memory until a corresponding one is
found. Once a representation is found, the connections relate it
to other representations. The pattern of connectivity in memory
reflects the topology of points in the environment. The cognitive
maps simply specify the possible "routes", they do not specify
which one to take. The specific "motivation" of the rat deter-
mines which route is selected. A motivation spreads through the
network as a signal that decreases from node to node: farther
nodes from the node first hit by the motivational signal will
reach a very weak signal. Action is determined by the motiva-
tional gradient on the map.
DeVega Manuel et al: MODELS OF VISUOSPATIAL COGNITION (Oxford
Univ Press, 1996)
A survey of theories on visual-spatial processing, mental
imagery, visual and propositional representations, etc.
Donald Merlin: ORIGINS OF THE MODERN MIND (Harvard Univ Press,
1991)
The book presents a theory of how the mind (symbolic thought)
arose from a nonsymbolic form of intelligence through gradual
absorbtion of new representational systems
Donaldson Margaret: CHILDREN'S MINDS (Norton, 1972)
A classic of developmental psychology, that expanded Piaget's
theory of different stages of mental development.
Donaldson Margaret: HUMAN MINDS (Penguin Press, 1992)
Donaldson provides a unifying vision of post-Piaget developmental
psychology (i.e., the growth of intellectual competence) by view-
ing the child's mental development as an organically growing
neural network shaped by the child's intentions. Piaget's stages
were defined by the ability to perform mental operations.
Donaldson's stages are defined by the child's focus of attention.
The first stage (first eight months of life), the "point mode",
is limited to things that the child can perceive directly ("here"
and "now"). The second stage, the "line mode", expands to
embrace the concepts of past and future ("there" and "then").
The third stage, the "construct mode" (second year of life) is
one of concern abut the nature of things ("anywhere and at any
time"). The fourth stage is the "transcendent mode", when the
child starts using its imagination ("nowhere").
The second part of the book delves into cultural history with far
less success.
Dougherty Ray: NATURAL LANGUAGE COMPUTING (Lawrence Erlbaum,
1994)
Prolog implementations of english, french and german grammars.
Dowty David: WORD MEANING AND MONTAGUE GRAMMAR (Reidel, 1979)
Drawing from the aristotelic classification of state, activity
and eventuality, Dowty thinks that the modal operators "do",
"become" and "cause" can be the foundations for building the
meaning of any other verb. A thematic role is a set of proper-
ties that are common to all roles that belong to that thematic
role. A thematic role being also a relationship that ties a term
with an event or a state, a Lambda calculus can be built on
thematic roles. Tematic roles are actually cognitive structures
that favor the acquisition of language. See Fillmore, Schank and
Jackendoff.
Dowty David: INTRODUCTION TO MONTAGUE SEMANTICS (Reidel, 1981)
One of the best books to understand Montague's thinking and prac-
tice. His intensional language is incrementally built starting
from truth-conditional, model-theoretic and possible-world
approaches to semantics, then introducing variables, quantifiers,
tense, modality and lambda calculus. The concepts underlying his
program for a "universal grammar" are also greatly simplified
and explained.
Drescher Gary: MADE-UP MINDS (MIT Press, 1991)
Drescher's "schema mechanism" is a computational implementation
of Piaget's theory of early child development. Concepts are built
through a stepwise process of synthesis and abstraction.
Dretske Fred: KNOWLEDGE AND THE FLOW OF INFORMATION (MIT Press,
1981)
Dretske is inspired by Shannon's and Weaver's theory of informa-
tion. In order to extend what is a purely "quantitative" theory
(dealing with the amount of information present in the state of a
system and the amount of information which is received in a
transmission between two systems) into a semantic theory of
information, Dretske distinguishes information from meaning (a
signal may have meaning but it certainly carries information) and
then relates information, knowledge and belief: knowledge as
information-caused belief (an agent knows that something is true
if having that information causes one to believe that it is the
case).
A state carries information about another to the degree that it
is lawfully dependent on that other state. The lawful relation-
ship between a cause and its effect accounts for the effect being
about the cause. Intentionality is not unique of mental states,
but quite ubiquitous in physical systems (for example, a thermom-
eter). Mental intentional states are somewhat limited compared to
physical systems' intentional states, as they miss a lot of
information that physical systems would not miss. In a sense, the
mind distorts the information that is available in the environ-
ment.
A state transports information about another state to the extent
that it depends on that state. Intentionality is reduced to a
cause-effect relationship: each effect refers to its cause.
There are systems outside the human mind which are intentional.
Having contents is not unique to the human mind, but having some
contents may be. What is unique is the transition from analogi-
cal information (as presented by sensors) to digital information
(the cognitive representation). Intentionality is "caused" by
the information perceived by the sensors. Coherently with
Gibson's and Neisser's theories, information is in the environ-
ment and cognitive agents simply absorb it, thereby creating men-
tal states.
The difference between sensory processes and cognitive processes
is reduced to the difference between analog processing and digi-
tal processing.
Perceptual systems are designed to maintain a stable correlation
between percept and the perceived world.
Intelligence is a function of the total capacity of information
processing.
A belief is a semantic structure whose content determines what is
believed. Beliefs require concepts and concepts imply the capa-
city for holding beliefs. A perceptual act creates a belief out
of a concept. A concept has both a backward-looking, informa-
tional aspect, and a forward-looking, functional aspect. What
concepts a system possesses is determined by the kind of informa-
tion to which its internal states are sensitive.
Similarly to Fodor ("narrow content" and "broad content" of a
mental representation) and Putnam (self-contained psychological
states such as pain versus world-related states such as "X loves
Y"), Dretske too has a two-factor theory of mental states: an
"indicator" (the "information", the causal relation to external
states) and a second factor which expresses the dependencies
between the internal states in a fashion reflecting the external
world.
Dretske Fred: EXPLAINING BEHAVIOR (MIT Press, 1988)
The term "behavior" is used in many different ways to mean dif-
ferent things. The behavior of an animal is commonly taken to be
the actions it performs more or less by instinct or by nature.
This is not necessarily "voluntary" behavior. The fact that women
have menstruations is part of "female behavior", but it is not
voluntary. Behavior is pervasive in nature, and cannot be res-
tricted to animals: plants exhibit behavior too. Behavior is the
production of some external effect by some internal cause.
Behavior is a complex causal process wherein certain internal
conditions produce certain external movements. First and
foremost, behavior is a process. A process is caused by both a
triggering cause (the reason why it occurs now) and a structural
cause (the reason why the process is the way it is). This holds
both for human behavior and the behavior of machines (a thermos-
tat switches on a furnace both because the temperature fell below
a threshold and because it has been designed to turn on furnaces
under certain conditions).
The explanation of purposive behavior in terms of intentions and
beliefs is not contradictory with a physical account of neural
and muscular activity. Generally, humans are interested in
structural behavior, which in plants and animals has been deter-
mined by natural evolution and in machines has been built by
humans.
The elements of a representational system have a content defined
by what it is their function to indicate (Grice's "non-natural
meaning"). Dretske distinguishes three types of representational
systems: Type I have elements (symbols) that show no intrinsic
power of representation (includes maps, codes, etc); Type II have
elements (signs) that are causally related to what they indicate
(includes gauges); Type III (or natural) have their own intrinsic
indicator functions (unlike Type I and Type II, in which humans
are the source of the functions) and therefore a natural power of
representation.
Dretske separates the reference of a representation from the
object that is causally responsible for the representation (a
gauge carries information about the item it is connected to, not
about which item it is that it is connected to)
In discussing the ccausal role of meaning, Dretske finds that the
intentional idiom of beliefs, desire, knowledge and intention can
as well be referred to primitive organisms that not only have a
system of internal structures whose relevance to the explanation
of behavior resides in what they indicate (they mean something
and mean something "to" the organism of which they are part).
Dretske Fred: NATURALIZING THE MIND (MIT Press, 1995)
Five lectures on consciousness, revolving around the thesis that
all mental facts are representational facts, which are in turn
facts about informational functions. What one thinks and feels is
determined by history and by the environment.
"Sense experience is the primary locus of consciousness".
Phenomenal experience dominates mental life. The phenomenal
aspects of perceptual experience are one and the same as external
real-world properties that experience represents objects as hav-
ing. Introspection is reduced to knowledge of internal facts via
an awareness of external objects. Sensations (seeing, smelling,
etc) are perceptual forms of consciousness.
Dretske provides an evolutionary account of sensory representa-
tion and ultimately of awareness. Animals that are conscious of
objects and events can do things in the environment that uncons-
cious animals cannot do.
Dretske Fred: SEEING AND KNOWING (University of Chicago Press,
1969)
Dretske believes that there are two fundamental versions of
vision: a non-epistemic seeing, that requires no belief in what
is being seen, and an epistemic seeing, which requires believing
in what is being seen. The object of the non-epistemic vision is
still a well-defined object, otherwise people who have no
knowledge of an object (or have different beliefs about that
object, such as an expert and a novice) would end up seeing dif-
ferent things when they look at it. In the epistemic mode, noth-
ing can be seen without first acquiring some true belief about
what is seen. This second way of seeing is subjective and may
vary considerably among individuals with different knowledge and
beliefs. Within epistemic seeing, a difference is drawn between
primary epistemic seeing (an object is identified in virtue of
how it looks) and secondary epistemic seeing (an object is iden-
tified not in virtue of the way it looks but in virtue of the way
other objects look with respect to it). A detailed mathematical
account of both ways of seeing is worked out.
Dreyfus Hubert: WHAT COMPUTERS CAN'T DO (Harper & Row, 1979)
The second edition of the book that started the anti-artificial
intelligence movement.
Inspired by Husserl's phenomenology (intelligence as a context-
determined, goal-directed activity), Dreyfus thinks that
comprehension can never do without the context in which it
occurs. The information in the environment is fundamental for a
being's intelligence. Dreyfus reviews ten years of research and
failures in artificial intelligence and proves that the four fun-
damental assumptions, biological (that the brain must operate as
a symbolic processor), psychological (that the mind must obey a
heuristic program), epistemological (that there must be a theory
of practical activity) grounds, and ontological (that the data
necessary for intelligent behavior must be discrete, expliciti
and determinate), are not plausible.
Dreyfus emphasizes the role of the body in intelligent behavior
and that human experience is intelligible only when organized in
terms of a situation (as a function of human needs).
The introduction to the second edition takes on Minsky's frames
and Schank's scripts, two noveties that apparently meet Husserl's
criteria for intelligence (in that they perform search for anti-
cipated facts). But they too assume that the context is a set of
rigidly defined situations, while in reality the context cannot
be separated from the rest of our everyday's lives.
Dreyfus Hubert & Dreyfus Stuart: MIND OVER MACHINE (Free Press,
1985)
A sobering critique of the foundations of artificial intelli-
gence, and more specifically symbolic problem solver.
Dreyfus claims that only novices behave like expert systems. The
expert has synthesized experience in an unconscious bahavior that
reacts istantaneously to a complex situation. What the expert
knows cannot be decomposed in rules.
The foundation of Dreyfus' argument is that minds do not use a
theory about the everyday world because there is no set of
context-free primitives of understanding. Human knowledge is
skilled "know-how", as opposed to expert systems' logical
representations, or "know-that".
Dubois Didier & Prade Henri: POSSIBILITY THEORY (Plenum Press,
1988)
The english translation of the original 1985 french text.
Possibility theory (formulated by Zadeh in 1977) developed as a
branch of the theory of fuzzy sets to deal with the lexical
elasticity of ordinary language (i.e., the fuzziness of words
such as "small" and "many"), and other forms of uncertainty which
are not probabilistic in nature. The subject of possibility
theory is the possible (not probable) values of a variable.
Imprecision is related to the value of an attribute of an object.
Uncertainty is related to the confidence in that value (probable,
possible, plausible, etc). Possibility theory is both a theory
of imprecision (represented by fuzzy sets) and a theory of uncer-
tainty. The uncertainty of an event is described by a pair of
degrees: the degree of possibility of the event and the the
degree of possibility of the contrary event. The definition can
be dually stated in terms of necessity, necessity being the com-
plement to one of possibility.
When the degrees of possibility can only take the value zero and
one, the calculus of possibility is identical to interval
analysis, in which imprecision is represented as sets of possible
values. Wuith continuous degrees of possibility those sets become
fuzzy sets.
The book introduces the mathematical tools of fuzzy logic.
Possibility logic (a logic of partial ignorance) extends modal
logic by assigning a degree of possibility and a degree of neces-
sity to each axiom.
Its basic axioms are that: 1. grade of possibility is one for a
proposition that is true in any interpretation and is zero for a
proposition that is false in any interpretation; 2. grade of pos-
sibility of a disjunction of propositions is the maximum grade of
the two. When the grade of necessity of a proposition is one,
the proposition is true. When the grade of possibility of a pro-
position is zero, the proposition is false. When the grade of
necessity is zero, or the grade of possibility is one, nothing is
known about the truth of the proposition.
Possiblity logic has a graded notion of possibility and neces-
sity, whereas in modal logic they are all-or-nothing concepts.
Possiblity logic admits only one set of axioms, while modal logic
admits many.
Dubois Didier, Prade Henri & Yager Ronald: READINGS IN FUZZY
SETS (Morgan Kaufmann, 1993)
All the historical papers from Lotfi Zadeh's 1965 "Fuzzy sets" to
Brat Kosko's "Adaptive inference in fuzzy knowledge systems". The
editors provide an intriguing survey of the prehistory of the
field, reaching back to Max Black and Karl Menger's "ensemble
flou". They also compare fuzzy logic with competing theories of
uncertainty, such as interval analysis and probabilities.
A few articles cover the foundations of fuzzy set theory. Dubois
and Prade discuss fuzzy numbers (fuzzy sets in the real line) and
possibility theory. Many articles cover applications to process
control and decision analysis.
Duchan Judith, Bruder Gall & Hewitt Lynne: DEIXIS IN NARRATIVE
(Lawrence Erlbaum, 1995)
Based on an interdisciplinary research program, the authors argue
in favor of a representational system (the "deictic center")
which readers construct when trying to understand a text by using
available knowledge. The deictic center contains temporal, spa-
tial and character information.
Dummett Michael: ELEMENTS OF INTUITIONISM (Oxford University
Press, 1977)
A general introduction to intuitionism. Intuitionism prescribes
that all proofs of theorems must be constructive. Only con-
structable objects are legitimate. The meaning of a statement
resides not in its truth conditions but in the means of proof or
verification.
Dummett Michael: TRUTH AND OTHER ENIGMAS (Harvard Univ Press,
1978)
The book collects many papers written by Dummett on various sub-
jects.
Dummett's theory of meaning is a variant of intuitionistic logic:
a statement can be said to be true only when it can be proven
true in a finite time (it can be "effectively decided", similar
to "intuitionistic justified"). In deciding truth one thing that
is required is understanding. A theory of meaning must explicit
what it is to know. A theory of meaning is an account of how
language is used. A theory of meaning is a theory of understand-
ing.
Dummett criticizes holism because it cannot explain how an indi-
vidual can learn language. If the meaning of a sentence only
exists in relationship to the entire system of sentences in the
language, it would never be possible to learn it. For the same
reason it is not possible to understand the meaning of a theory,
if its meaning is given by the entire theory and not by single
components.
Dummett Michael: SEAS OF LANGUAGE (Clarendon, 1993)
A collection of many articles about philosophy of language from
the point of view of his theory of meaning.
Dyson Freeman: INFINITE IN ALL DIRECTIONS (Harper & Row, 1988)
A physicist's speculations on the origins of life and on the
relationship between science and faith.
Life is metabolism and replication, and they are separable.
Therefore it is possible that life began twice, first with
creatures capable of metabolism and then with creatures capable
of replication. Dyson argues that the fundamental characteris-
tics of life must be homeostasis (rather than replication),
diversity (rather than uniformity), the cell (rather than the
gene). The origins of life must be consistent with life's macros-
copic features: looseness of structure and tolerance of errors.
Dyson also speculates on the connection between cosmology and
biology. Inspired by Jamal Islam, who calculated how matter
would evolve in universes which expand forever, Dyson calculates
mathematically what life is and how it will evolve. A closed
universe is doomed to collapse and life with it. Since a system's
entropy is a measure of the number of alternative states of the
system, the complexity of a living organism should be propor-
tional to the negative of its entropy. Dyson even computed the
entropy of a human being (the rate at which humans dissipate
energy times the human body's temperature times the duration of a
unit of consciousness): 10 to the 23th.
Eccles John: EVOLUTION OF THE BRAIN (Routledge, 1991)
The book offers a history of human evolution, of the evolution of
the hominid brain, of the evolution of speech production, of evo-
lution of visual skills, of evolution of learning and memory. A
key role is assigned to the limbic system and, in general, to the
latest evolutionary additions to the human brain, the cerebral
neocortex.
Then Eccles delves into a study of the evolution of conscious-
ness. Drawing from Margenau, Eccles argues that the mind-brain
interaction is analogous to a probability field of quantum
mechanics. Mental "energy" can cause neural events by a process
analogous to the way a probability field causes action. He calls
"psychon" the mental unit that transmit mental intentions to the
neural units.
From a detailed analysis of the cerebral neocortex, Eccles
derives that cerebral asymmetry is a fundamental property of the
human brain, that the self is unique to the left hemisphere, and
that the neo-neocortex is the site of gnostic functions. Cons-
ciousness resides in a psycological world that transcends the
physical. The soul is a separated entity from the body, and is
created by God.
Eccles John: THE SELF AND ITS BRAIN (Springer, 1994)
The anti-materialist view of this book focuses on a spiritual
self that is capable of controlling the materic brain and bring-
ing about voluntary movement.
Edelman Gerald: NEURAL DARWINISM (Basic, 1987)
Gerald Edelman is possibly the main contributor to the selec-
tional theory of the immune system. When the body is attacked by
a virus, it produces specially adapted protein molecules, antibo-
dies, that attach themselves to the invaders and destroy them.
Those antibodies are created by the thousands "before" the body
is attacked by anything. An invasion results in a rapid increase
in the rate of production of the one antibody that matches the
intruder. Edelman is now applying the same concept to a selec-
tional theory for brain development, thereby introducing popula-
tion thinking to neurobiology.
Before birth the genetic instructions in each organism provide
general constraints for neural development, but cannot specify
the exact location and configuration of each cell. After birth
innate "values", i.e. adaptive cues (such as "looking for
food"), generate behavior and therefore feedback from the
environment, which in turns helps "select" the neural configura-
tions that are more suitable for survival. During this on-going
process of "learning" the brain develops categories by selec-
tively strengthening or weakening connections between neural
groups. Experience "selects" one configuration of neural groups
out of all the configurations that are possible.
The functioning of the brain can be explained as resulting from a
morphological selection of neural groups. Neural groups "compete"
to respond to environmental stimuli. Each brain is therefore dif-
ferent, depending on the stimuli that it encounters during its
development.
Adhesion molecules determine the initial structure of neural
groups, the "primary repertory". Behavior determines the secon-
dary repertory. Repertories are organized in "maps", each map
having a specific neural function. A map is a set of neurons in
the brain that has a number of links to a set of receptor cells
or other maps.
Maps communicate through parallel bidirectional channels, i.e.
the "reentrant" signaling. Reentry is not just feedback because
there can be many parallel pathways operating simultaneously.
The process of reentrant signaling allows a perceptual categori-
zation of the world, i.e. to relate independent stimuli. This
ability enables higher level functions such as memory.
In Edelman's view brain processes are dynamic and stochastic.
The brain is not an "instructional" system but a "selectional"
system. It evolves not by changes in a constant set of neurons
but by selection of the most valuable neural groups among those
that exist since birth. And the elementary unit of this process
is not the single neuron, but the neural group.
Edelman Gerald: TOPOBIOLOGY (Basic, 1988)
The title refers to location-dependent development of body cells:
how can a cell know where in the body it is supposed to grow in
order to generate the shape and function of the animal?
Edelman's molecular embryology claims that development is based
on topobiological events (division, movement, death and so forth
of cells, which are regulated by cell-adhesion and substrate-
adhesion molecules on the surface of the cell). A cell's com-
petence is due essentially to its location.
Animate systems exhibit three properties that allow them to
exist: heredity, variation in their hereditary material, competi-
tion as the environment changes. Animate systems are self-
replicating systems, whose genetic code undergoes mutation and
whose variant individuals undergo natural selection.
Characteristic of animate systems is development, in particular
morphogenesis, the emergence of form during embryonic develop-
ment. Roughly the same cell types appear in different parts of
the body. The difference in position and shape results from the
interaction of a number of driving forces (namely cell division,
cell motion and cell death), which determine the number of cells
in a particular region, and regulatory processes (namely cell
adhesion and cell differentiation), which determine the interac-
tion among cells.
Evolution can be viewed as a process of phenotypic transformation
resulting largely from genetically mediated change in
developmental dynamics that is itself altered throughout phylo-
geny.
Edelman than analyzes in detail what he considers the molecular
mechanisms of epigenesis.
Development is under genetic control, but developmental events
are nonetheless epigenetic and topobiologically controlled. Pat-
tern, and not mere cell differentiation, is the evolutionary
basis of morphogenesis. The cell surface, not its core, plays the
fundamental role in this process, because it mediates signals
from other cells and links with other surfaces to form tissues.
A sequence of interactions between certain special types of genes
via epigenetic signal paths provides the basis of pattern by con-
trolling temporal sequences of mitosis, movement, death and
further signaling.
In order to explain how this process can be reconciled with
extensive changes in animal form in relatively short evolutionary
time periods, Edelman points to the nonlinear relation between
genetics, development and evolution.
Edelman Gerald: THE REMEMBERED PRESENT (Basic, 1989)
Edelman's biological theory of consciousness begins with his
theory of how higher-level cognitive functions emerge: from reen-
trant processes. Consciousness arises from the interaction of
two parts of the neural system that differ in their anatomical
structure and evolutionary history: the one responsible for
categorizing (external stimuli) and the one responsible for
"instinctive" behavior (homeostatic control of behavior). At
this level concepts are not absolute, but can be remembered.
"Primary consciousness" (being aware of things in the world)
therefore arises from "reentrant loops" that interconnect "per-
ceptual categorization" and "value-laden" memory. Primary cons-
ciousness has an evolutionary reason to be, since it helps
abstract and organize complex changes in the environment.
In order to have higher consciousness the brain must also be able
to make the distinction between the self and the rest of the
world and to order events in time. A higher-level consciousness
(being aware of itself), unique to humans, is then possible if
the brain is capable of: perceptual categorization, memory,
learning and self-nonself discrimination.
Edelman thinks that two parts of the nervous system differ radi-
cally in their evolution, organization and function. And that
consciousness emerges as the product of an ongoing categorical
comparison of the workings of those two kinds of nervous system.
The part that is crucial to consciousness has evolved to be dedi-
cated to adaptive, homeostatic and endocrine functions related to
the individual's immediate needs for survival. Such functions
therefore reflect evolutionarily selected values that have con-
tributed to fitness. Regions that are assigned to define self
within a species include the amygdala, the hippocampus, the lim-
bic system, the hypothalamus. Regions that operate to define non-
self include the cortex, the thalamus and the cerebellum.
From an evolutionary point of view, the milestone moment was when
a category-value link emerged, because then the basis for cons-
ciousness was laid.
Edelman then provides a detailed neurophysiological model of how
memory works, in particular how time and space (and successions
within them) are represented can be represented by brain organs.
Edelman thinks that concept formation preceded language. Concepts
are driven by the perceptual system and stored in memory. With
the advent of language concepts become absolute, independent of
time. The brain structures that are responsible for concept for-
mation are those that can categorize, discriminate and recombine
patterns of activity in different kinds of global mappings.
Language was enabled by the evolutionary emergence of special
anatomy: the acquisition of phonological capacities provided the
means forst for semantics and then for syntax to arise by linking
the preexisting conceptual learning with the emerging lexical
learning.
Edelman Gerald: BRIGHT AIR BRILLIANT FIRE (Basic, 1992)
The book summarizes Edelman's theory of neural development and
consciousness formation. In practice, Edelman extends an account
of the development of perceptual categories into a general
account of consciousness.
The reentry mechanism between maps yields a process of "global
mapping" that leads to the creation of perceptual categories and
generalization. Edelman distinguishes between primary conscious-
ness (imagery and sensations) and higher-order consciousness
(language and self-consciousness). Primary consciousness
requires memory (a process of both storing and recategorizing),
value (a way to rank stimuli and eventually to learn), discrimi-
nation of the self from the non-self, a way to represent chronol-
ogy, and global reentrant pathways connecting all these struc-
tures. Higher-order consciousness.
Edelman thinks that science cannot solve the problem of qualia
because no two people will have the same qualia.
Eigen Manfred & Schuster Peter: THE HYPERCYCLE (Springer Verlag,
1979)
The origin of life from inorganic matter is due to emergent
processes of self-organization.
Hypercycles are a class of nonlinear reaction networks that can
originate spontaneously within the population of a species
through natural selection and naturally evolve to higher complex-
ity by allowing for the coherent evolution of a set of function-
ally coupled self-replicating entities. Natural selection itself
is inevitable: given a set of self-reproducing entities that feed
on a common and limited source of energetic/material supply,
natural selection will spontaneously appear.
A hypercycle is based on nonlinear autocatalysis (reproduction
cycles which are linked by cyclic catalysis, i.e. by another
autocatalysis). A hypercycle is therefore the next higher level
in the hierarchy of autocatalytic systems.
The second part of the book analyses the behavior and mathemati-
cal properties of hypercycles.
The model explains the simultaneous unity (due to the use of a
universal genetic code) and diversity (due to the "trial and
error" approach of natural selection) in evolution. This dual
process started even before life was created. Evolution of
species was preceded by an analogous stepwise process of molecu-
lar evolution.
Systems can be classified in four groups according to their sta-
bility with respect to fluctuations: stable systems (the
fluctuations are self-regulating), indifferent systems (the fluc-
tuations have no effect), unstable systems (self-amplification of
the fluctuations) and variable systems (the system can show
either regulation, indifference or amplification of fluctua-
tions). Only the last type (indifference towards a broad mutant
spectrum, stability towards selective advantages and instability
towards unfavorable configurations) is suitable for generation of
biological information. Selection is a mathematical consequence
of the dynamics of self-reproducing systems of this kind.
Eigen's experiments with RNA proved that under suitable condi-
tions a solution of nucleotides give rise spontaneously to a
molecule that replicates, mutates and competes with its progeny
for survival. The replication of RNA appears to be the fundamen-
tal event around which the rest of biology developed. First
genes were created, then proteins, then cells. Cells simply pro-
vide physical cohesion.
Eigen Manfred: STEPS TOWARDS LIFE (Oxford University Press,
1992)
By employins his "hypercycle" technique, Eigen speculates how
living cells and bodies may have come to be, starting from molec-
ular tools: cells first learned to self-replicate and then to
surround themselves with protective membranes.
Eigen also uncovers a feedback mechanism inherent in natural
selection that favors (or accelerates the search for) superior
mutants. This explains the apparently impossibly fast rate of
adaptation by viruses. That feedback mechanism turns what would
be a steady function of improvement into an exponential function
of improvement, thereby explaining how viruses can adapt so
quickly. The feedback mechanism is due to the fact that the "wild
type" of a genotype (the pure genotype) is always surrounded by
almost identical variants, and this accelerates the emergence of
superior mutants.
Ekman Paul & Davidson Richard: THE NATURE OF EMOTION (Oxford
Univ Press, 1994)
A series of articles on emotion from psychologists.
Engelmore Robert: BLACKBOARD SYSTEMS (Academic Press, 1988)
All the historical papers on the subject, from Barbara Hayes-Roth
to Nii. Opportunistic planning was first used in the HEARSAY
system in the mid Seventies, then formalized in 1979 by Frederick
and Barbara Hayes-Roth ("A cognitive model of planning").
Hayes-Roth's opportunistic and incremental model of reasoning
contemplates many independent agents cooperating to find the
solution to a problem. Each specialized agent is triggered by
information written by other agents on a blackboard and each
agent can in turn write information for other agents on that
blackboard.
The system keeps two agendas, one for the actions it "wishes" to
perform (those that at least one agent needs to continue its rea-
soning) and one for the actions that it "can" perform (those
whose preconditions have been satisfied). By matching necessary
and possible actions the system determines which agents are
active at any time. The computational advantage of this model of
inference is that only actions that are relevant to the solution
of the problem are taken into consideration.
Epstein Richard: SEMANTIC FOUNDATIONS OF LOGIC (Kluwer Academic,
1990)
A general introduction to the most popular varieties of proposi-
tional logics. Epstein sets himself to defining his "relatedness
logic", a logic which takes into account the subject matter of
propositions, and "dependency logic", which, similarly, focuses
on the referential content of a proposition.
A broad coverage of modal logics (and Kripke's semantics), intui-
tionism (Brouwer's manifestos, Heyting's formalization and
Kripke's semantics), many-valued logics (Lukasiewicz, Post,
Kleene) is also provided.
Epstein Richard: SEMANTIC FOUNDATIONS OF LOGIC: PREDICATE LOGIC
(Kluwer Academic, 1994)
A vast, technical introduction to predicate logic, semantics,
identity, quantifiers, descriptive names, functions, second-order
logic.
Estes William: CLASSIFICATION AND COGNITION (Oxford University
Press, 1994)
Estes offers a psychological theory of memory organization based
on categorization. Estes distinguishes classification (partition-
ing a set of objects in a set of groups) and categorization (par-
titioning plus each category implies a set of properties for its
members). After an historical overview, Estes advances his core
model, a combination of an array framework (in which memory
interfaces with perception by means of a mechanism based on simi-
larity and in which the association between memory and action
varies according to a learning mechanism) and the product rule
(by which similarity of two patterns is computed as a product of
the differences between each pair of corresponding features of
the two patterns). Basically, Estes adopts both a storage-
retrieval model and an adaptive network model, thereby marrying
cognitive psychology and connectionism.
A system of categorization based on the product rule differs
considerably from prototype-based systems such as Rosch's.
Estes' model is based on empirical data and provides a rigorous
mathematical formulation.
Eysenck Michael: PRINCIPLES OF COGNITIVE PSYCHOLOGY (Lawrence
Erlbaum, 1993)
A short introduction to the field.
Fauconnier Gilles: MENTAL SPACES (MIT Press, 1994)
A revised edition of the 1985 cognitive linguistics classic that
described how discourse constructs mental spaces. Mental spaces
are domains that are built by the hearer as she listens to a
speech. They are interconnected and consist of elements, roles,
strategies and relations between them. Fauconnier applies the
theory to presuppositions and counterfactuals.
Feigenbaum Edward: COMPUTERS AND THOUGHT (MIT Press, 1995)
A collection of articles by Turing, Newell, Simon, Minsky,
Feigenbaum, etc.
Feigl Herbert: THE MENTAL AND THE PHYSICAL (Univ of Minnesota
Press, 1967)
In this 1957 essay Feigl argues in favor of the class identity
theory of the mind. Physical and mental terms may have different
senses but identical referents: mental states may refer exactly
to the same states as do physical states, even if they describe
the states in a completely different manner. Mental idioms and
physical idioms are different descriptions of the same states.
Mental states and physical states have the same extension but
different intension: they describe the same states, but in a dif-
ferent way.
In the postscript to the second edition Feigl rejected his origi-
nal theory and opted for eliminativism: there is no evidence of a
relation between mental and physical states, and only the physi-
cal (neuroscientific) language should be employed in discussing
people's feelings.
Fetzer James: ASPECTS Of ARTIFICIAL INTELLIGENCE (Kluwer, 1988)
A collection of philosophical articles on machine intelligence,
notably Fetzer's own introduction to the theory of semiotic sys-
tems. Newell's and Simon's hypothesis of the mind as a symbol
processing system can be extended by considering the mind as a
semiotic system, i.e. sign processing systems. Fetzer thinks
that symbol systems simulate mental processes that semiotic sys-
tems replicate.
Fetzer James: ARTIFICIAL INTELLIGENCE (Kluwer, 1990)
Fetzer thinks that the standard model of Artificial Intelligence,
that views minds as symbol processing systems, is fundamentally
flawed, because minds are semiotic systems. Fetzer introduces to
the theory of semiotic systems. The notions of semantic net-
works, frames, scripts are reviewed in the philosophical context
of a theory of knowledge, belief and action.
Fetzer James: EPISTEMOLOGY AND COGNITION (Kluwer, 1991)
A collection of philosophical papers (mainly critiques) that deal
with Fodor's computational theory of the mind, connectionism,
scripts, frames and so forth.
Fiesler Emile & Beale Russell: HANDBOOK OF NEURAL COMPUTATION
(Oxford Univ Press, 1996)
The ultimate handbook for professional neural network designers.
It includes applications to Biology, Medicine, Economics, etc.
Finke Ronald: PRINCIPLES OF MENTAL IMAGERY (MIT Press, 1989)
A survey of psychological findings about mental imageries. Finke
identifies five principles of equivalence between a mental
imagery and the perceived object: the principle of implicit
encoding (informatin about the properties of an object can be
retrieved from its mental image), the principle of spatial
equivalence (parts of a mental image are arranged in a way that
corresponds to the way that the parts of the physical object are
arranged), the principle of perceptual equivalence (similar
processes are activated in the brain when the objects are ima-
gined as when they are perceived), the principle of transforma-
tional equivalence (imagined transformations and physical
transformations are governed by the same laws of motion), the
principle of structural equivalence (the mental imagery exhibits
structural features corresponding to those of the perceived
object such that the relations between the object's parts can be
both preserved and interpreted).
Finke Ronald: CREATIVE IMAGERY (Lawrence Erlbaum, 1990)
A book devoted to the psychological phenomenon that people can
detect emergent patterns in imagery even if they were not aware
of them when the image was formed. Most of these recognitions
occur only when people inspect their images.
Finke Ronald: CREATIVE COGNITION (MIT Press, 1992)
A study of creativity in terms of the cognitive processes and
structures that make it possible. The model includes a genera-
tive phase, in which mental representations (or "preinventive"
structures) that promote creative discovery, and an exploratory
phase, in which they are interpreted in meaningful ways.
Finke Ronald: CHAOTIC COGNITION (Lawrence Erlbaum, 1996)
Chaotic thinking is the process by which the individual copes
with a world full of unpredictability, changes and uncertainties.
Fisher Ronald Aylmer: THE GENETICAL THEORY OF NATURAL SELECTION
(Dover, 1929)
Seminal work that highlighted how genes from the parents are
reshuffled in each new generation. Fisher used sophisticated
mathematics in dealing with evolution, thereby providing a scien-
tific account of how a distribution of genes in a population will
change as a result of natural selection.
Fisher erred in thinking about the evolution of the single gene,
neglecting the influence of all the other genes, and in assuming
that evolution was a process of achieving stable equilibrium.
Flanagan Owen: CONSCIOUSNESS RECONSIDERED (MIT Press, 1992)
Flanagan's book is an introduction to the issues concerning cons-
ciousness: qualia, self-consciousness, memory, sensations and
multiple personalities disorders. It does not provide a model to
explain what consciousness arises from, but it examines the
phenomena that may lead to such an explanation. Consciousness is
considered as a natural phenomenon that can be explained by sci-
ence.
Flanagan Owen: THE SCIENCE OF THE MIND (MIT Press, 1991)
Descartes' dualism violates the principle of conservation of
energy. William James' work is the first formulation of the
naturalistic position in the philosophy of mind: the mental is
physical, although it cannot be explained by mechanical laws, and
it has an evolutionary purpose; consciousness is not an entity,
but a function. Flanagan reviews Freud's psychoanalysis,
Skinner's behaviorism, Piaget's and Kohlberg's theories of cogni-
tive development, the main themes of Cognitive Science and Artif-
icial Intelligence, and Wilson's sociobiology.
Consciousness is an heterogeneous set of processes which have in
common the property of being felt. Flanagan does not believe in
"one" consciousness, but in a group of "conscious" phenomena.
Some of the processes of our body are unconscious and non per-
ceived (the heartbeat), some are unconscious but perceived by
other processes (sensors), and some are conscious, perceived by
themselves.
Flanagan Owen: CONSCIOUSNESS RECONSIDERED (MIT Press, 1992)
A review of phenomena related to consciousness, from qualia to
multiple personalities.
Flanagan Owen: SELF EXPRESSION (Oxford Univ Press, 1996)
A series of essays on subjects related to consciousness, dreams,
and psychological disorders.
Flood Raymond & Lockwood Michael: NATURE OF TIME (Basil
Blackwell, 1986)
A collection of essays about the arrow of time (time's inherent
directionality, in spite of the apparent symmetricity of the fun-
damental laws of nature) and the second law of thermodynamics
(the only law of nature which is not symmetric).
Penrose's "Big bangs, black holes and time's arrow" deals with
the apparent contradiction of increasing entropy in an universe
that started in a state of maximum entropy (thermal equilibrium
before the big bang) and in an universe whose fundamental laws
are all symmetric.
Paul Davies relates the direction of time to the quantum collapse
of the wave function. Davies also suggests that the mind-body
problem may be related to quantum mechanics' dualism between
waves and particles, as the mind's role (of information encoding
and processing) is similar to the wave's role.
Dummett's "Causal loops" refutes all arguments against the possi-
bility that we can influence our past.
Fodor Jerry: LANGUAGE OF THOUGHT (Crowell, 1975)
Fodor's computational theory of the mind views the mind as a spe-
cial symbolic processor. Propositional attitudes can be explained
by assigning a symbolic memory to each possible attitude (hope,
desire, fear, etc) and and each symbol to one of the possible
propositions. A proposition in an attitude constitutes a proposi-
tional attitude. Each symbol is a "mental representation" and the
mind is endowed with a set of rules to operate on such represen-
tations. Cognitive life is the transformation of those rules.
Mental representations constitute a language of thought, "men-
talese".
Evidence of an internal language in the mind comes from rational
behavior (the ability to compute the consequences of an action),
concept learning (the ability to form and verify hypotheses) and
perception (the ability to recognize an object or an event).
These phenomena would not be possible if the agent was not able
to represent to itself the elements of the problem.
Such language cannot be one of the languages we speak because the
very ability to speak requires the existence of an internal
language of representation.
But the language of thought exhibits features that are shared by
human languages: productivity (ability of understanding and pro-
ducing propositions from an infinite set by using recursive
operations over finite resources), systematicity (a physical
relation between mental representations so that one can yield
others), coherence (ability to make syntactically and semanti-
cally plausible inferences).
The mind processes symbols without knowing what those symbols
mean, in a purely syntactic fashion. Behavior is due only to the
internal structures of the mind.
All knowledge is represented syntactically.
Fodor Jerry: REPRESENTATIONS (MIT Press, 1981)
A collection of philosophical essays on the representational
theory of the mind.
Fodor looks for an explanation of how propositional attitudes can
have semantic properties. Propositional attitudes are relations
(between an agent and a state of the world). Among the relata are
mental representations. Mental representations are symbols,
endowed with both syntactic and semantic properties. They possess
their causal role in virtue of their syntactic properties. Propo-
sitional attitudes inherit their semantic properties from the
mental representations that function as their objects.
Fodor Jerry: MODULARITY OF THE MIND (MIT Press, 1983)
Fodor advances a theory of the mind that exsumes Gall's view of
vertical faculties. Cognitive faculties can be divided in verti-
cal faculties (which are domain-specific, genetically determined,
computationally autonomous and associated with distinct neural
structures) and horizontal faculties. Modular cognitive systems
are vertical faculties. systems of input analysis and systems
that subserve the fixation of belief.
Fodor Jerry: A THEORY OF CONTENT (MIT Press, 1990)
A collection of papers on Fodor's theory of mental content.
Fodor speculates that there exist two types of meaning. Fodor
discriminates between "narrow content" and "broad content" of a
mental representation: the former is a semantic representation,
is purely mental and does not depend on anything else; the latter
is a function that yields the referent in every possible world,
and depends on the external world.
Meaning is the ordered set of narrow and broad contents. Narrow
content is a conceptual role. As in Sellars, a role is a purely
syntactic property, as they occur in formal systems.
Fodor claims that there is no type identity but only instance
identity. Mental instances that constitute a mental class can be
used by neural events which do not form a neural class.
Fodor Jerry & Lepore Ernest: HOLISM (Basil Blackwell, 1992)
The book is a critical survey of the theory that only whole
languages or whole belief systems really have meanings; and the
meanings of smaller units are merely derivative. Each chapter
attacks the thinking of an influential philosopher: Quine, David-
son, Lewis, Dennett, Block and Churchland.
Fodor's "rational fixation" of beliefs is a non-demonstrative
process that employs analogy and induction.
Fodor Jerry: THE ELM AND THE EXPERT (MIT Press, 1994)
A lively introduction to the issues of the mental language plus a
critique of the critique of his theory. His opponents' claim that
referential semantics cannot provide a robust theory of inten-
tional explanation is rebuffed by positing that psychological
laws are intentional, psychological processes are computational
and the semantic properties of mental representations are
referential (semantics is purely informational).
Forbus Kenneth & DeKleer Johan: BUILDING PROBLEM SOLVERS (MIT
Press, 1993)
A textbook that focuses on truth maintenance systems.
Forrest Stephanie: EMERGENT COMPUTATION (MIT Press, 1991)
A collection of papers on the topic of emergent computation.
Most papers assume that physical systems exist that can support
computation, and analyze under which conditions computational
processes may come to be spontaneously.
Emergent computation is to standard computation what nonlinear
systems are to linear systems: it deals with systems whose parts
interact in a nontrivial way.
Chris Langton presents his theory of computation ad the edge of
chaos: physical systems achieve the prerequisites for the emer-
gence of computation (i.e., transmission, storage, modification)
in the vicinity of a phase transition. Specifically, information
becomes an important factor in the dynamics of cellular automata
in the vicinity of the phase transition between periodic and
chaotic behavior. In that neighborhood, information can propagate
over long distances without decaying appreciably, thereby allow-
ing for long-range correlation in behavior (ordered configura-
tions do not allow for information to propagate at all, and
disordered configurations cause information to quickly decay into
random noise). This conclusion is consistent with Von Neumann's
findings. A fundamental connection is therefore displayed
between computation and phase transition.
Kauffman debates orderly dynamics and frozen components as
requirements for the evolvability of complex systems. He also
notes how nonlinear dynamical systems which interact with the
external world classify and know their world through their
attractors.
Holland, as well as Forrest, looks at emergent computation in
classifier systems. Hillis proves that co-evolving parasites
help improve evolution.
A number of papers deal with connectionism. Daniel Greening sur-
veys a variety of parallel simulated annealing techniques.
Churchland views explanatory understanding, perceptual recogni-
tion and abductive inference as different instances of prototype
activation.
Franklin Stan: ARTIFICIAL MINDS (MIT Press, 1995)
An excellent interdisciplinary survey of artificial intelligence,
cognitive science, artificial life, neurobiology. Franklin
presents recent theories of the mind by Chalmers, Sloman, Grif-
fin, Minsky, Ornstein; describes the SOAR cognitive architecture,
Brooks' subsumption architectures, Brustoloni's autonomous
agents, Drescher's schemata, Kanerva's sparse distributed memory,
Edelman's neural darwinism, Maturana's autopoiesis. discusses
Dreyfus' and Penrose's critiques of artificial intelligence;
introduces the theory of dynamic systems.
Frost Richard: INTRODUCTION TO KNOWLEDGE BASED SYSTEMS (MacMil-
lan, 1986)
A comprehensive introduction on how we can build systems that are
capable of storing and processing complex pieces of knowledge.
Notions and techniques from database technology, formal logic,
expert systems research and advances in natural language (each of
which are discussed at length in a very scientific manner) are
linked to yield the foundations of a complete and unified theory
of knowledge representation.
Frost covers many-sorted logics, non-monotonic logic, many-valued
logics (including fuzzy logic), modal logics (alethic, deontic,
epistemic), the main variants of temporal logic, the theory of
types, Montague's intensional logic and theories of uncertainty
(probability, possibility, plausibility)
Then Frost delves into knowledge representation techniques: pro-
duction rules, semantic networks, frames, scripts and formalizes
the types of inference that they enable.
Functional language is described, with emphasis on the Lambda
calculus,
Throughout the book a rigorous mathematical notation is employed.
Gallistel C.R.: THE ORGANIZATION OF ACTION (Erlbaum, 1980)
The nature of intelligence lies in the organization principles
that enable living organisms to make rapid adjustments of pat-
terns of action in response to the environment. No movement in
nature is random, it always serves the purpose of "adapting" the
state of the system to the external conditions. No matter how
intelligent a living being's action appears to be, that action
satisfies the same general principle. The reason human actions
look more complex than the actions of inanimated matter is
because of the complexity of the human machine, i.e. of the
brain's neural circuitry. The subtleties of goal, intent, pur-
pose are but consequences of the hierarchical synthesis of inter-
mediate units.
The elementary units of behavior (reflex, oscillator, ser-
vomechanism, i.e. externl stimulus to internal signal to muscle
contraction) are "catalyzed" by units at the higher levels of the
system. Gallistel describes the interaction principles that
govern the units of behavior (reciprocal facilitation, reciprocal
inhibition, chaining, superimposition, acceleration/deceleration,
corollary discharge, etc). The goal is to explain how an action
that looks like a whole can be decomposed in many coordinated
lower-level levels.
Drawing from Paul Weiss' concept of a central program, Gallistel
assumes that units are organized in a hierarchy that allows for
competition and antagonism. A central program is a unit of
behavior that is activated as a whole. A central program "selec-
tively potentiate" subsets of lower-level units according to
their relevance to the current goal. The principles that deter-
mine the "selective potentiation" of lower-level units are the
same that govern the properties of elementary units.
Drawing from Deutsch's theory of learning, which prescribes how
representations of the world determine action, Gallistel defines
cognition as the representation of the world stored in memory.
Gallistel therefore argues in favor of innate knowledge, i.e.
universal principles of behavior.
The book contains reprints of historical papers (Sherrington's
study of the reflex, Von Holst's oscillators, Wilson's on coordi-
nation, Fraenkel's analysis of geotaxis) and a wealth of experi-
mental data.
Galton Antony: TEMPORAL LOGICS (Academic Press, 1987)
Six essays from authoritative researchers in the field of tem-
poral logic. Galton provides an overview of both the first-order
(Davidson, McDermott, Allen, Kowalski) and the modal (Prior's)
approaches. Sadri discusses in detail Kowalski's calculus of
events, Lee's logic of time and events, Allen's temporal logic.
Galton presents his logic of occurrence
In his logic of aspect an event-radical is a complete expression
that is neither a proposition nor a name, but it denotes an event
type. Occurrences are event tokens: each single occurrence of an
event type is an occurrence. Aspect operators (perfect, progres-
sive and prospective) are applied to event-radicals to yield pro-
positions. Such operators express the occurrence of events in
time. The logic of occurrence is the logic of such operators.
Galton Antony: THE LOGIC OF ASPECT (Clarendon Press, 1984)
"Aspect" refers to the fact that every verb has two forms, the
imperfective (used to describe an action in progress) and perfec-
tive (used to describe a completed action). Aspect is related to
tense: aspect determines how tense has to be interpreted (e.g.,
perfective aspect is incompatible with present tense).
Prior worked out a logic of tenses. Galton extends that logic by
introducing a distinction between events (which are perfective)
and states (imperfective). States "obtain" in moments, whereas
events "occur" in intervals. Aspects are treated like operators.
Prior's two temporal operators are still applied to states to
obtain new states but two new operators transform events into
states and two more transform states into events.
Gamut L.T.F.: LOGIC, LANGUAGE AND MEANING (University of Chi-
cago, 1990)
J. Benthem, J. Groenendijk, D. De Jongh, M. Stokhof and H. Ver-
kuyl provide a broad introduction to the standard and intensional
logics, pragmatics and Montague's grammar.
Gardner Howard: MIND'S NEW SCIENCE (Basic, 1985)
A history of cognitive research, that spans cybernetics, neuro-
physiology (Lashley, Hebb), philosophy of the mind (Ryle,
Wittgenstein, Austin), psychology (Miller, James, Kohler,
Bartlett, Piaget), artificial intelligence, linguistics, anthro-
pology, biology (Gibson, Marr).
Gardner Howard: FRAMES OF MIND (Basic, 1983)
Gardner argues that there is no single, unified, indivisible
intelligence, but rather a set of independent intellectual com-
petences. Gardner finds the biological foundations of intelli-
gence in the plasticity of the neural system during development.
Garnham Alan & Oakhill Jane: THINKING AND REASONING (Blackwell,
1994)
A cognitive psychology approach to inference (deduction and
induction), creativity, common sense and the development of
cognition.
Gazdar Gerald: PRAGMATICS (Academic Press, 1979)
Pragmatics studies aspects of meaning that cannot be accounted
for by reference to truth conditions. Pragmatics deals with mean-
ing minus truth conditions, or meaning minus semantics. In his
approach to the field Gazdar employs a formalist methodology
analogous to the one applied By Montague to semantics.
Gazdar offers a critique of the theory of illocutionary force
based on the performative hypothesis (that the deep structure of
every sentence contains a performative verb).
After recapitualting Grice's treatment of implicatures and four
maxims, Gazdar proposes to replace the quality maxim with "say
only that which you know", so that implicatures due to the maxim
of quality (both scalar implicatures and clausal implicatures)
can be treated as Hintikka's epistemic implications, thereby
solving the "projection problem" (how the presuppositions of a
sentence are determined by those of its components).
After a reasoned critique of existing treatments of presupposi-
tion (Hausser, Katz, Langendoen, Stalnaker, Karttunen), Gazdar
offers his definition, drawing from Hamblin's "commitment store"
model of dialogue and Bar-Hillel's view of an utterance as the
pair of a sentence and a context. Gazdar offers an inductive
definition of context (a set of propositions constrained only by
consistency) and uses Stalnaker's pragmatic definition of sen-
tence meaning.
Gazdar Gerald: GENERALIZED PHRASE STRUCTURE GRAMMAR (MIT Press,
1985)
Gazdar abandons the transformational component and the deep
structure of Chomsky's model of grammar and focuses on rules that
analyze syntactic trees rather than generate them. They translate
natural language sentences in an intensional logic which is a
variant of lambda calculus.
Gazdar's grammar describes only context-free languages and exhi-
bits mathematical properties that, unlike Chomsky's grammar, can
be scientifically tested and falsified. A phrase-structure rule
is not a generative rule but a condition of compliance for a syn-
tactic tree. The semantic interpretation of a sentence is derived
directly from its syntactic representation.
Gazdar defines 43 rules of grammar each providing a phrase-
structure rule and a semantic-translation rule that shows how to
build an intensional-logic expression from the intensional-logic
expressions of the constituents of the phrase-structure rule.
Gazdar employs meta-rules to produce new rules (and therefore
derived categories) from the existing rules.
Gazzaniga Michael & LeDoux Joseph: INTEGRATED MIND (Plenum,
1978)
Based on the results of split brain experiments, the authors
present a theory that what is transferred between the emispheres
is neural codes to maintain an informational balance and provide
for mental unity.
The authors criticize the view that the two emispheres are highly
specialized units, and reduce lateralization to the lateraliza-
tion of linguistic skills.
Gazzaniga Michael: SOCIAL BRAIN (Basic, 1985)
Humans are more of a sociological entity than a single unified
psychological entity. The human brain is social.
Gazzaniga's model of the brain (and the mind) is modular:
independent units work in parallel. A special module is the
"interpreter" of behavior, which makes sense ex-post of even the
most capricious acts. Beliefs are created by the interaction of
the interpreter with the other modules. Gazzaniga looks for evi-
dence of his theory in neurophysiology, archaeology and anthro-
pology.
Gazzaniga Michael: NATURE'S MIND (Basic, 1992)
Gazzaniga emphasizes that innate factors play a key role in
determining human behavior. Following Edelman, brains are born
with a vast number of pre-wired circuits, but most often they
offer many alternative options for development. Experience
determines which of these pre-existing brain circuits are used.
Many possible connections can be made, but only some are selected
by experience.
Genesereth Michael & Nilsson Nils: LOGICAL FOUNDATIONS OF ARTIF-
ICIAL INTELLIGENCE (Morgan Kaufman, 1987)
A textbook on Artificial Intelligence that covers production sys-
tems (predicate calculus, deduction, resolution), some nonmono-
tonic logics (closed-world assumption, circumscription, default
theory), inductive learning, probabilistic reasoning, logics of
belief, and planning. The last chapter attempts to define an
intelligent agent at three levels: a tropistic agent, that simply
reacts to the environment; a hysteretic agent, that has an inter-
nal state; and a knowlegde-level agent, whose internal state is
basically determined by a production system.
Gell-Mann Murray: THE QUARK AND THE JAGUAR (W.H.Freeman, 1994)
A book on complexity (i.e., nonlinearity) that tries to bridge
the simple (e.g., elementary particles) and the complex (e.g., a
living organism).
"It is not simple to define simple". Gellman defines it as the
absence of complexity.
According to superstring theory subatomic particles are compacti-
fied hyperdimensional space (matter having originated when six of
the original dimensions of space collapsed into superstrings).
Gell-man provides a modern account of quantum mechanics, based on
Richard Feynman's view of many alternative possible histories of
the universe as a direct consequence of chance. The probabilis-
tic nature of quantum mechanics allows the universe to unfold in
an infinite number of ways. The second law of thermodynamics per-
mits the temporary growth of order in relatively isolated,
energy-driven systems.
Complex adaptive systems behave in accordance with the second law
of thermodynamics. Biological evolution is a complex adaptive
system that complies with that law once the entire environment,
and not only the single organism, is taken into account. Once
complex adaptive systems establish themselves they operate
through a cycle that involves variable schemata, randomness,
phenotypic consequences and feedback of selection pressures to
the competition among schemata.
Living organisms dwell "on the edge of chaos", as they exhibit
order and chaos at the same time, and they must exhibit both in
order to survive. Living organisms are complex adaptive systems
that retrieve information from the world, find regularities,
compress them into a schema to represent the world, predict the
evolution of the world and prescribe behavior for themselves. The
schema may undergo variants that compete with one another. Their
competition is regulated by feedback from the real world under
the form of selection pressure. Disorder is useful for the
development of new behavior patterns that enable the organism to
cope with a changing environment.
Gibson James Jerome: THE SENSES CONSIDERED AS PERCEPTUAL SYSTEMS
(Houghton Mifflin, 1966)
Gibson originated "ecological realism", the view that meaning is
located in the interaction of living things and the environment.
Perceiving is a process of picking up information that is avail-
able in the environment. Perception is a constant process and
consists in detecting the invariants. The function of the brain
is to orient the organs of perception for seeking and extracting
information from the continous energy flow of the environment.
Perception cannot be separated from the environment in which the
perceptive system evolved and from the information which is
present in that environment. There is much more information in
the world and less in the head than was traditionally assumed.
The environment must be viewed as a source of stimulation.
Conscious sensation and perception are two different things and
they are often independent. Perceptual systems are sources of
information. Sensations are sources of conscious qualities. The
inflow of information does not always coincide with the inflow of
sensations. Therefore, a study of sensations is not very useful
to a study of perceptions.
Perceptual organs are not passive. They can orient themselves to
pick up information, to "resonate" with the information in the
environment. Gibson goes to a great length to explain the details
of their functioning.
Gibson James Jerome: THE ECOLOGICAL APPROACH TO PERCEPTION
(Houghton Mifflin, 1979)
According to Gibson the correct context for a theory of action is
not the abstract space of objects and their relationships but the
real world of shapes and colors as it is presented by the senses.
Perception and action are not separate processes. Organisms move
in the world using all the information that is available in it.
Information originates from the interaction between the organism
and its environment.
An "affordance" measures conjunctions between the characteristics
of the organism and the environment. All the potential uses of an
object constitute the activities it affords (e.g. a pen affords
writing). Such uses are directly perceivable.
Ginsberg Matthew: READINGS IN NONMONOTONIC LOGIC (Morgan Kauf-
mann, 1987)
Ginsberg introduces the problems that led to the development of
nonmonotonic logics, first and foremost property inheritance by
default (such as "Tweety is a bird" implies that "Tweety can
fly"). Then he classifies formal approaches to nonmonotonic
inference: proof-based approaches (Raymond Reiter's default
logic), modal approaches (Drew McDermott's nonmonotonic logic,
Robert Moore's autoepistemic logic) and minimization approaches
(John McCarthy's circumscription).
Reiter's "A logic for default reasoning" (1980) introduces the
"closed world axiom" (what is not true is false), or negation as
failure to derive (if a ground atomic formula cannot be proved
using the premises, then assume the formula's negation), then
defines his default logic with the following inference rule: "if
A is true and is consistent that B is true, then assume that B is
also true" (or "if a premise if true, then the consequence is
also true unless a condition contradicts what is known").
McDermott's formulation of modal logic (1980) is based on a
coherence operator ("P is coherent with what is known" if P can-
not be proven false by what is known).
Moore's "Semantical consideration of nonmonotonic logic" (1985)
removed some of the problems with McDermott's with his "autoep-
istemic logic", based on the notion of belief (related to
McDermott's coherence), which extends Hintikka's epistemic modal
logic to incorporate action. The logic models the beliefs of an
agent reflecting upon his own beliefs. Moore also provides a
possible-world semantics for his logic.
In 1987 Kurt Konolige proved that autoepistemic and default logic
are formally identical.
Yoav Shoham (1987) argues that all approaches to nonmonotonic
reasoning can be reduced to a partial ordering on the set of all
models for a theory.
Ginsberg argues that a variety of approaches to nonmonotonic rea-
soning can be unified by resorting to multi-valued logics.
Jon Doyle's "Truth Maintenance System" (1979) was the first
effective computational frameworks for default reasoning. It
consists of a problem solver that draws inferences and a system
that records those inferences (or "justifications"). It main-
tains beliefs and justifications for beliefs. It ensures that
the database is free of contradictions by identifying and adding
justifications to remove contradictions whenever they are
discovered (dependency-directed backtracking).
Johan DeKleer (1986) improved the concept with his "assumption-
based" TMS which labels each proved proposition with the sets of
premises needed to derive it (its "context").
In 1986 McDermott introduced the "temporal projection problem"
(which occurs when trying to infer which facts are true once a
sequence of events have occurred) and proved that none of the
nonmonotonic approaches can deal with it. The logic should
select not the minimal models, but the chronologically minimal
models.
Shoham's "Chronological ignorance" (1986) formalizes the idea of
chronological minimization by temporally ordering the conflicting
extensions that underlie it and preferring the later ones (those
in which abnormality occurs as late as possible).
By employing possible worlds, Ginsberg's "Reasoning about action"
(1987) solves the frame, ramification and qualification problems
and circumvents the temporal projection problem. As is Richard
Fikes' STRIPS, a single model of the world is updated when
actions are performed by constructing the nearest world to the
current one in which the consequences of the actions under con-
sideration hold. The nearest world is found by constructing
proofs of the negation of the explicit consequences of the
expected action and by removing a premise in each proof from the
current world.
The book contains McCarthy's "Some philosophical problems from
the standpoint of Artificial Intelligence" (1969), "Epistemologi-
cal problems of Artificial Intelligence" (1977), and "Cir-
cumscription"" (1980).
The first article is the one that introduced situation calculs
and the frame problem. The second identifies the qualification
problem and the third one details his theory of circumscription.
According to McCarthy, knowledge representation must satisfy
three fundamental requirements: ontological (must allow one to
describe the relevant facts), epistemological (allow one to
express the relevant knowledge) and heuristic (allow one to per-
form the relevant inference). Artificial Intelligence can be
defined as the discipline that studies what can be represented in
a formal manner (epistemology) and computed in an efficient
manner (heuristic). The language of logic satisfies those
requirements: it allows us to express everything we know and it
allows us to make computations on what is expressed by it. Each
set of knowledge is in fact a mathematical theory.
McCarthy's situation calculus represents temporally limited
events as "situations" (snapshots of the world at a given time),
by associating a situation of the world (set of facts that are
true) to each moment in time. Actions and events are functions
from states to states. An interval of time is a sequence of
situations, a "chronicle" of the world. The history of the world
is a partially ordered sequence of states and actions. The pro-
perty of states is permanence, the property of actions is change.
Each situation is expressed in a formula of first-order predicate
logic. Causal relations between two situations can then be com-
puted. A state is expressed by means of logical expression that
relate objects in that state. An action is expressed by a func-
tion that relates each state to another state.
McCarthy's "frame problem" states that it is not possible to
represent what does not change in the universe as a result of an
action, as that is an infinite set of facts. Complementary para-
doxes are the "ramification problem" (infinite things change
because one can go into greater and greater detail of descrip-
tion) and the "qualification problem" (the number of precondi-
tions to an action is also infinite).
Circumscription deals with default inference by minimizing abnor-
mality: an axiom that states what is abnormal is added to the
theory of what is known (predicate circumscription). The objects
that can be shown to have a certain property, from what is known
of the world, are all the objects that satisfy that property (or,
the only individuals for which that property holds are those
individuals for which it must hold). This definition involves a
second-order quantifier. This is analogous to Frege's method of
forming the second-order definition of a set of axioms: such a
definition allows both the derivation of the original recursive
axioms and an induction scheme stating that nothing else satis-
fies those axioms.
Glass Leon & Mackey Michael: FROM CLOCKS TO CHAOS (Princeton
University Press, 1988)
The authors propose nonlinear models for dynamic processes occur-
ing in body organs (biological oscillators).
Gleick James: CHAOS (Viking, 1987)
The best seller that made chaos theory fashionable. Besides
exposing the theory in ordinary language, and highlighting its
applications to many different disciplines, it provides a pic-
turesque chronicle of the field. Chaos theory is about finding
regularities in the irregular behaviors of nature, i.e. in the
behavior of nonlinear systems. Chaotic systems are a subset of
nonlinear systems in which small changes in initial conditions
yield big changes in behavior.
Glezer Vadim: VISION AND MIND (Lawrence Erlbaum, 1995)
Starting from a description of how the visual system works,
Glezer develops a detailed neural theory of how categories are
formed from sensory inputs through functional organization of
neural structures.
The mind somehow models the external world. Comprehension of the
world is achieved through functional modules in the neocortex,
whose first task is to segment the sensory input. This is done in
a way consistent with Gabor's quantum theory of information
(indeterminacy between the description of signals in space and in
spatial-frequency domains, i.e. duality between space and spec-
trum). The harmonics of each module have different properties (a
Fourier analysis is provided).
An invariant representation of the object is localized in the
left emisphere, i.e. the left emisphere uses a classification
approach for recognition, whereas the right emisphere uses a
structural approach. Invariance emerges as a property of neural
nets and Hebbian learning (an algorithm for the production of
invariant representations of an image is provided).
Gluck Mark & Rumelhart David: NEUROSCIENCE AND CONNECTIONIST
THEORY (Lawrence Erlbaum, 1990)
A collection of articles on how brain regions can be modeled to
account for function, complexity and power.
Goddard Cliff & Wierzbicka Anna: SEMANTIC AND LEXICAL UNIVERSALS
(Benjamins, 1994)
A collection of papers on the theme of Leibniz's universal alpha-
bet of thought, the set of semantic and lexical universals that
are supposed to be common to all languages. At the end Wierz-
bicka gives a critical account of all the primitives (37 of them)
that have been identified. "Canonical" sentences are those con-
structed out of such primitives.
Goldberg David: GENETIC ALGORITHMS (Addison Wesley, 1989)
This is the book that explained what Holland's theories were all
about. Goldberg defines genetic algorithms as "search algorithms
based on the mechanics of natural selection and natural genet-
ics". Unlike most optimization methods, that work from a single
point in the decision space and employ a transition method to
determine the next point, genetic algorithms work from an entire
"population" of points simultaneously, trying many directions in
parallel and employing a combination of several genetically-
inspired methods to determine the next population of points.
Goldberg focuses on efficiency issues and possible applications.
Simple algorithms such as reproduction (that copies chromosomes
according to a fitness function), crossover (that switches seg-
ments of two chromosomes) and mutation are discussed, as well as
more complex algorithms such as dominance (a genotype-to-
phenotype mapping), diploidy (pairs of chromosomes) and abeyance
(shielded against overselection); inversion (the primary natural
mechanism for recoding a problem, by switching two points of a
chromosome); and many micro-operators.
A classifier system is a machine learning system that learns syn-
tactically rules (or "classifiers") to guide its performance in
the environment. A classifier system consists of three main com-
ponents: a production system, a credit system (such as the
"bucket brigade") and a genetic algorithm. Goldberg verifies an
important properties of Holland's classifiers: the trend to
create "standard hierarchies", in which a general rule covers
normal situations but many exception rules take over in those
situations where the default rule would not work.
Goldfield Eugene: EMERGENT FORMS (Oxford Univ Press, 1995)
The book offers a theory of how functional acts (such as eating,
walking, smiling) emerge during infancy: through the assembling
of a variety of biodynamic devices. The theory can be extended to
cognition and language.
Goodwin Brian: HOW THE LEOPARD CHANGED ITS SPOTS (Charles
Scribner, 1994)
The organism, and not the gene, should be the focus of attention
for evolutionary biologists. Goodwin argues in favor of a theory
of morphogenesis as a process that is inherently ordered. Genes'
instructions are constrained by a principle of order.
Gould Stephen Jay: ONTOGENY AND PHYLOGENY (harvard University
Press, 1977)
Gould reviews the debate on "recapitulation", the idea that onto-
geny (individual development) recapitulates phylogeny (species
development) and advances a theory that views "heterochrony"
(changes in developmental timing that producing parallels
between ontogeny and phylogeny) as evolutionary crucial. Retarda-
tion (delayed growth and development), for example, has probably
been fundamental for the evolution of humans, by prolonging into
later life rapid brain growth and therefore an increase in cere-
bralization.
Gould Stephen Jay: EVER SINCE DARWIN (Deutsch, 1978)
An accessible introduction to darwinism, neo-darwinism and
Gould's own theory of punctuated equilibria (changes appear sud-
denly in lineages) and non-repeatability of evolution (if evolu-
tion had to happen again, it would not repeat itself).
The paleontological record shows no steady progress in the
development of higher organisms. Evolution seems to proceed in
bursts.
Consciousness is probably the latest burst of evolutionary
development.
Gould also touches on theories of the earth and the nature of
science.
Gould Stephen Jay: WONDERFUL LIFE (Norton, 1989)
A popular introduction to the significance of the findings of the
Burgess Shale. Gould advances intriguing hypotheses: any replay
of the tape of life would yield a different, unpredictable evolu-
tionary history, but still a meaningful one. Evolution is not in
the hands of determinism and not in the hands of randomness, but
in the hands of contingency. In the case of the creatures of the
Burgess shale, survival was so unlikely that chance events may
well have shaped evolution more than fitness. Humans exist
because of a lucky chain of events that led to them, but they
might have as well never been created.
Gould Stephen Jay: FULL HOUSE (Random House, 1996)
Gould shatters stereotypes about evolution by claiming that bac-
teria represent the dominant form of life. Gould restates his
point that evolution does not proceed towards complexity but ran-
domly produces variety. Progress is purely accidental. With
Darwin, there is only variation, not progress. Gould reiterates
his theory of punctuated equilibrium: mostly nothing happens.
but when it happens, it happens quickly. Consciousness evolved
only once in all the experiments life performed on Earth (whereas
eyes evolved several dozen times, and wings even more often).
Consciousness is therefore unlikely to occur, and human cons-
ciousness must be considered a sheer accident. If the tape of
life were played back again, it is unlikely that a conscious
being would emerge. On the other hand life may be more probable
than it appears to be: it happened on the Earth as soon as it
could happen.
Graf Peter & Masson Michael: IMPLICIT MEMORY (Lawrence Erlbaum,
1993)
A technical introduction to the field of explicit and implicit
memory. Each chapter is written by an expert in the field.
Implicit memories are those in which experiences influence per-
formance in the absence of specific intention to recollect them.
Graubard Stephen: THE ARTIFICIAL INTELLIGENCE DEBATE (MIT Press,
1988)
A collection of more or less philosophical articles on the feasi-
bility of Artificial Intelligence. Hubert and Stuart Dreyfus draw
from Heidegger and Wittgenstein to affirm their conception of a
holistic intelligence, that cannot be broken down into knowledge
representation systems or neural networks, of an intelligence
that is driven by intentions which reflect the environment. Put-
nam even downplays the historical importance of Artificial Intel-
ligence.
The book also contains introductions to current research in con-
nectionist models, machine vision, etc.
Green David: COGNITIVE SCIENCE (Blackwell, 1996)
A textbook on cognitive science that covers the history and the
main topics of this discipline in a conversational style.
Green Georgia: PRAGMATICS (Lawrence Erlbaum, 1989)
Pragmatics is defined as the study of understanding intentional
human action. Therefore it must deal with belief, goal, plan and
act. Green surveys indexical and anaphoric expressions (expres-
sions whose reference cannot be determined without taking into
account the context, such as pronouns and demonstratives, whose
interpretation requires inferences about the speaker's intended
referent), sense and reference (Frege's distinction of extension
and intension, Kripke's and Putnam's casual theory of names,
Kripke's distinction of rigid designators and non-rigid designa-
tors in the context of possible worlds, and Montague's inten-
sional logic in which the sense of an expression is supposed to
determine its reference)
Green deals at length with illocutionary force (what action an
utterance is performing) and presupposition (the facts that are
taken for granted), two linguistic phenomena without which many
utterances could not be evaluated. Grice's maxims are presented
as the basis to assess the coherence of a discourse. Metaphor is
treated as something different from implicature (indirect
speech), but similar in terms of the strategies that must be
employed to understand it.
Greene Robert: HUMAN MEMORY (Lawrence Erlbaum, 1992)
A comprehensive survey of cognitive models from the perspective
of cognitive psychology.
Gregory Richard: OXFORD COMPANION TO THE MIND (Oxford, 1987)
A monumental, detailed, accurate reference book. An alphabetical
dictionary of mental phenomena and brain anatomy, spanning
psychology, philosophy and neurophysiology. Each entry is writ-
ten by experts in the field and reviews the state of the art on
the subject.
Grice H. Paul: STUDIES IN THE WAY OF WORDS (Harvard Univ Press,
1989)
Grice thinks that language is based on a form of cooperation
among the speakers. People always choose the speech acts that
achieve the goal with minimum cost and highest efficiency.
Grice was influential in emphasizing the linguistic interplay
between the speaker, who wants to be understood and cause an
action, and the listener. This goes beyond syntax and semantics.
A sentence has a timeless meaning, but also an occasional mean-
ing: what the speaker meant to achieve when s/he uttered it.
Language has meaning to the extent that some conventions hold
within the linguistic community. Those conventions help the
speaker achieve his/her goal. The participants of a conversation
cooperate in saying only what makes sense in that circumstance.
The significance of an utterance includes both what is said (the
explicit) and what is implicated (the implicit). Grice therefore
distinguishes between the proposition expressed from the proposi-
tion implied, or the "implicature". Implicatures exhibit proper-
ties of cancellability (the implicature can be removed without
creating a contradiction) and calculability (an implicature can
always be derived by reasoning under the assumption that the
speaker is observing pragmatic principles). A particularized
implicature is one that is such in virtue of the context. A gen-
eralized implicature is independent of the context.
Grice's four maxims summarize those conventions. They help the
speaker say more than it says through implicatures which can be
implied by the utterance. Conventional implicatures are deter-
mined by linguistic constructions in the utterance. Conversa-
tional implicatures follow from maxims of truthfulness, informa-
tiveness, relevance and clarity that speakers are assume to
observe. Conversational implicatures can be discovered through an
inferential process: the hearer can deduct that the speaker meant
something besides what he said by the fact that what he said led
the hearer believe in something and the speaker did not do any-
thing to stop him from thinking it.
The maxims are: provide as much information as needed in the con-
text, but no more than needed (quantity), tell true information
(quality), say only things that are relevant to the context
(relation), avoid ambiguity as much as possible (manner).
Griffin Donald: ANIMAL THINKING (Harvard University Press, 1984)
A study of animals' minds. Griffin claims that smaller brains
have a greater need to think because they can store fewer infor-
mation. The only way they can cope with their environment is by
thinking more.
Grishman Ralph: COMPUTATIONAL LINGUISTICS (Cambridge, 1986)
An introduction to the field (syntax, semantics, discourse
analysis and language generation) that provides detailed discus-
sions of various parsing techniques, a brief discussion on ana-
phora resolution, a survey of frames and scripts.
Grossberg Stephen: NEURAL NETWORKS AND NATURAL INTELLIGENCE (MIT
Press, 1988)
The book explores a number of phenomena and proposes a potential
explanation in terms of neural dynamics.
Grossberg's connectionist model, consistent with Hebb's law and
Pavlov's conditioning, reduces a cognitive state to a dynamic
state of "adaptive resonance", expressed by a non-linear, non-
stable and non-local algorithm.
The essential element of the cognitive system is the long-lasting
state of adaptive resonance reached when the feedback matches the
input pattern. Inspired by Helmholtz, Grossberg thinks that we
perceive the sensory data only when a consensus is reached
between what the data are and what we expect them to be, given
what we already know. This competetive negotiation is reached
through a feedback process. Grossberg's theory of memory is non-
linear, nonlocal and nonstationary.
Grosz Barbara: READINGS IN NATURAL LANGUAGE PROCESSING (Morgan
Kaufman, 1986)
Focusing mainly on discourse analysis and historical natural
language systems, it contains seminal papers by Perrault, Hobbs,
Grosz, Wilensky.
Perrault touches on mathematical properties of some popular
linguistic formalisms: transformational grammars, Jerrold
Kaplan's and Joan Bresnan's lexical-functional grammars, Gerald
Gazdar's generalized phrase-structure grammar and Joshi's tree
adjunction grammar.
Woods's 1970 paper describes a parser for non context-free
languages: augmented transition networks, an extensions of recur-
sive transition networks (directed graphs with labeled arcs and
nodes) that have registers for storing partial parse trees or
flagging features and conditions for testing registers to deter-
mine how to proceed. ATN's are equal in capacity to transforma-
tional grammars.
In 1980 Fernando Pereira and David Warren introduced a formalism
(definite clause grammar) to define grammars based on Horn-
clauses: grammar rules are written as logical formulas. Parsing
is a form of theorem proving.
Coming to semantics, Drew McDermott's "No notation without deno-
tation" played the role of a manifesto for systematic semantics:
it is not only important that a syste be correct, it is also
important that it can be understood.
Grosz thinks that dialogues exhibit a structure much like sen-
tences and this structure affects the use of referring expres-
sions. This "intentional" structure is characterized by a global
focus of attention and a number of immediate foci of attention.
Grosz analyzes the relationship between focus of attention and
referring expressions. Grosz defines a focus space as that sub-
set of the speaker's total knowledge which is relevant to a
discourse segment. Several such spaces may be relevant at a time.
A number of papers cover the integration of natural language
understanding and planning techniques (although Richard Fikes'
STRIPS is not mentioned). Raymond Perrault and his associates
model speech acts as operators and intentions as plans. Wilensky
uses plan recognition for story understanding.
Historical natural language processing systems from the Seventies
reviewed here include William Woods' LUNAR, Daniel Bobrow's GUS
and Terry Winograd's SHRDLU.
Gupta Anil & Belnap Nuel: THE REVISION THEORY OF TRUTH (MIT
Press, 1993)
According to Gupta's revision theory of truth, originally formu-
lated in the early 80's, truth is a circular concept. Therefore
all paradoxes that arise from circular reasoning in classical
logic fall into normality in Gupta's theory of truth.
In Gupta's "revisionist theory of truth" truth is refined step by
step. In order to determine all the sentences of a language that
are true when that language includes a truth predicate (a predi-
cate that refers to truth), one needs to determine whether that
predicate is true, which in turn requires one to know that the
extension of true is, while such extension is precisely the goal.
The solution is to assume an initial extension of "true" and then
gradually refine it.
Haken Hermann: SYNERGETICS (Springer-Verlag, 1977)
Synergetics is a theory of pattern formation in complex systems.
It tries to explain structures that develop spontaneously in
nature.
Since order emerges out of chaos, and chaos is not well defined,
synergetics employs probabilities (to describe uncertainty) and
information (to describe approximation). Entropy becomes a cen-
tral concept, relating physics to information theory.
Synergetics revolves around the concepts of: compression of the
degrees of freedom of a complex system into dynamic patterns that
can be expressed as a collective variable; behavioral attractors
of changing stabilities; and the appearance of new forms as none-
quilibrium phase transitions.
Systems at instability points are driven by a slaving principle:
long-lasting quantities can enslave short-lasting quantities
(i.e., they can act as order parameters). Close to instability,
stable motions (or "modes") are enslaved by unstable modes and
can be ignored, thereby reducing the degrees of freedom of the
system. The macroscopic behavior of the system is determined by
the unstable modes. The dynamic equations of the system reflect
the interplay between stochastic forces ("chance") and deter-
ministic forces ("necessity").
Synergetics deals with self-organization, how collections of
parts can produce structures. Synergetics therefore applies to
systems driven far from equilibrium, where the classic concepts
of thermodynamics are no longer adequate. Order can arise from
chaos and can be maintained by flows of energy/matter.
Applictions to Physics, Chemistry, Sociology and Biology (popula-
tion dynamics, evolution, morphogenesis) are discussed. Com-
pletely different systems exhibit surprising analogies as they
pass through an instability.
Biological systems are unique in that they exhibit and interplay
between structure and function. Function is embodied in struc-
ture, function is latent in form.
Synergetics belongs to the class of mathematical disciplines
(including Von Bertalanffi's general systems theory and
Prigogine's nonequilibrium thermodynamics) that are trying to
extend science to dynamic systems.
Hamblin Charles: IMPERATIVES (Basil Blackwell, 1987)
The book describes Hamblin's action-state semantics for dealing
with imperatives. The theory provides for a time scale, distinc-
tion between actions and states, physical and mental causation,
agency and action-reduction, and intensionality.
Hamilton Ternell: PROCESS AND PATTERN IN EVOLUTION (MacMillan,
1967)
Mutation, recombination, selection and isolation are the driving
forces of evolution. Natural selection results in differential
reproduction, i.e. in adaptation of populations, i.e. in evolu-
tionary change. The phenotype of an organism is the result of the
conflict between different selection forces. The individual is
the unit of natural selection, gene substitution is the unit pro-
cess in adaptation, and the species is the major unit of evolu-
tion.
Hamilton thinks that evolution is accelerated by parasites.
Organisms adopted sexual reproduction in order to cope with inva-
sions of parasites. Life is a symbiotic process which necessi-
tates of competitors.
Hampson Peter & Morris Peter: UNDERSTANDING COGNITION
(Blackwell, 1995)
An introduction to the main topics of cognitive psychology:
memory, vision, language, attention. Three paradigms for studying
cognition are discussed: artificial intelligence, cognitive sci-
ence and connectionism.
Hanson Norwood: PATTERNS OF DISCOVERY (Cambridge Univ Press,
1958)
We see what we know. In order to see what another person sees we
first need to learn what he knows. As we learn new knowledge, the
world as we perceive it changes.
Harris MaryDee: INTRODUCTION TO NATURAL LANGUAGE PROCESSING
(Prentice Hall, 1985)
An excellent textbook on how to process natural language with a
computer. It starts with a historic review, from Chomsky to
Fillmore's case grammar and generative semantics. The main
chapters address transformational generative grammar (phrase
marker, transformational rules, etc); transition networks (recur-
sive and augmented); case grammar; semantic networks; Schank's
conceptual dependency; knowledge representation (scripts,
frames).
Hassoun Mohamad: FUNDAMENTALS OF ARTIFICIAL NEURAL NETWORKS (MIT
Press, 1995)
A textbook on neural networks that begins with linear threshold
gates, expands computational properties into the most popular
supervised and unsupervised learning rules. A neural network is
defined as a parallel computational model comprised of densely
interconnected adaptive processing units in which learning by
example replaces programming.. Neural learning is viewed
mathematically as a search/approximation method. Extensive treat-
ment is provided of adaptive multilayer networks. The book makes
an effort to provide a unified and logical summary of the field.
Hassoun Mohamad: ASSOCIATIVE NEURAL MEMORIES (Oxford, 1993)
Articles by James Anderson, Pentti Kanerva, Amir Dembo and lots
of japanese contributions.
Haugeland John: ARTIFICIAL INTELLIGENCE (MIT Press, 1985)
An introduction to the field, that begins with an overview of
modern science and explains the basic concepts for a broad audi-
ence.
Hayes-Roth Frederick: BUILDING EXPERT SYSTEMS (Addison Wesley,
1983)
Haykin Simon: NEURAL NETWORKS (Macmillan, 1994)
One of the most comprehensive and updated surveys of neural net-
work algorithms.
Hebb Donald: ESSAY ON MIND (Lawrence Erlbaum, 1980)
Hebb's cell-assemblies theory holds that repeated exposure to a
sensory stimulation will result in an assembly. Thought processes
consist of an activity of such cell-assemblies. There is an inti-
mate relationship between learning and perception: perception in
the early stages is consequence of a primitive learning pro-
cess, but later learning becomes a function of perception and of
cognitive structure that originate from perception.
Hebb also makes a few philosophical comments. The word "cons-
cious" is used both for denoting the state of a human being and
for denoting a type of mental activity. The idea of the self and
the idea of the other overlap, and this explains the existence of
empathy.
Hebb Donald: THE ORGANIZATION OF BEHAVIOR (John Wiley, 1949)
Hebb's hypothesis is that the basis for neural development lay in
a selective strengthening or inibition of synapses between neu-
rons. Synapses that get used are reinforced, while synapses that
are not used are inhibited. This dual process molds the structure
of the brain in a darwinian fashion. Metabolic change therefore
occurs in the brain all the time. These synaptic changes are the
basis for all learning and memory.
Besides advancing the learning rule for synaptic modification,
the books defines the notion of the brain as a connectionist dev-
ice and the notion that within the brain regions of intercon-
nected self-reinforcing subnets of neurons (or "cell assemblies")
form for long periods of time.
The brain is an evolutionary system: genes determine only its
initial configuration, experience molds the brain according to
darwinian principles of selection.
The selective strenghtening of the synapses causes the brain to
organize itself into cell assemblies, each assembly representing
a fragment of a concept, each assembly overlapping others so that
concepts are naturally linked into larger concepts. Each resonat-
ing cell assembly behaves like a rule: triggered by an event,
will fire for a while at a higher rate.
Psychological conditioning is ubiquitous in animals because it is
a property of individual neurons.
Hecht-Nielsen Robert: NEUROCOMPUTING (Addison-Wesley, 1989)
A textbook on neural networks (parallel, distributed, adaptive
information processing systems), from a pragmatic, industrial
viewpoint. All the most popular learning laws are examined
extensively.
Hecht-Nielsen uses Kolmogorov's theorem to demonstrate that for
every function there exists a three-layer neural net which can
compute its values.
Heil John: PERCEPTION AND COGNITION (Univ of California Press,
1983)
Heil attempts to reconcile Gibson's theory of perception, that
perception is largely a process of gathering of information from
the environment, with a cognitive account of cognition. Percep-
tion is a link between beliefs and events or objects. In the end
perception is the acquisition of beliefs by way of the senses.
Concepts are simply skills that enable the perceiving agent to
acquire beliefs. Having beliefs does not necessarily require
language. Having beliefs does not necessarily require internal
representations or computational capabilities.
The class of perceptual objects for a perceiving agent is deter-
mined by 1. the agent's sensory system (which is sensitive to
some environmental stimuli and not others, and even for those
stimuli it is tuned to detect only some high-order features) and
2. the agent's set of concepts, or perceptual beliefs.
Heil has modified Dretske's theory by assuming, with Kant, that
the transition from analogic to digital is made possible by con-
cepts that are innate in the agent.
Herbert Nick: ELEMENTAL MIND (Dutton, 1993)
Herbert thinks that consciousness is a pervasive process in
nature. Mind is as fundamental a component of the universe as
elementary particles and forces. Mind can be detected by three
quantum features: randomness, thinglessness (objects acquire
attributes only once they are observed) and interconnectedness
(John Bell's discovery that once two particles have interacted
they remain connected). Herbert delves into psychological, parap-
sychological and even mystic phenomena that are supposed to cor-
roborate his hypotheses. Herbert reviews models of awareness
based on quantum effects.
Hertz John, Krogh Anders & Palmer Richard: INTRODUCTION TO THE
THEORY OF NEURAL COMPUTATION (Addison-Wesley, 1990)
A textbook on neural networks that starts with the Hopfield model
and then covers perceptrons, multi-layer networks, Boltzmann
machines, unsupervised learning (adaptive resonance, Kohonen).
It provides a very modern expositions of the computational con-
cepts.
Hewitt Carl: TOWARDS OPEN INFORMATION SCIENCE (MIT Press, 1990)
Hewitt has developed a semantics of intelligent communities. A
system is "open" when the outcome of its actions can be predicted
and at any time it can absorb new information from the outside
world. Distributed intelligent systems are a particular type of
open systems that can interact. The dynamics of such systems
depends on the balance between two factors: self-reliance, i.e.
the ability to act based only on local resources, and interdepen-
dency, the need to find resources elsewhere. That translates
into the dualism of "committment" (the action that a system is
determined to perform) and "cooperation" (the set of mutually
dependent roles among systems).
The main property of such systems is their "deductive indecisive-
ness": since many agents compete for the same resources in paral-
lel, the state of the world at any time is indeterminate. The
distributed system can only exhibit "global coherence".
Heyting Arend: INTUITIONISM (North Holland, 1956)
A classic textbook for intuitionism. Intuitionism prescribes
that all proofs of theorems must be constructive. Only con-
structable objects are legitimate. The meaning of a statement
resides not in its truth conditions but in the means of proof or
verification.
Hintikka Jaakko: KNOWLEDGE AND BELIEF (Cornell Univ Press, 1962)
A very technical epistemic and doxastic theory. Hintikka sets up
a formal system and shows its applications to the use of the
verbs "know" and "believe".
Hintikka Jaakko THE INTENTIONS OF INTENTIONALITY (Reidel, 1975)
A collection of articles, including "Objects of knowledge", which
defines the principles of his logic of attitudes.
Propositional attitudes can be interpreted using possible worlds
and an "alternativeness" relation. Alternatives are relative to
an attitude, an agent and the world in which the agent has that
attitude. The sentence "a believes that p" can be therefore
interpreted as "a believes that p is true in a world if and only
if p is true in all the alternatives to that world".
Following Gibson's biological theory, Hintikka argues that per-
ception is intentional because it is informational. Possible-
world semantics is advanced as a promising candidate for a gen-
eral theory of intentionality.
Hintikka Jaakko: THE GAME OF LANGUAGE (Reidel, 1983)
Hintikka proposed his "game-theoretical semantics" as an alterna-
tive to compositional semantics. The semantic interpretation of a
sentence is conceived of as a game between two agents. The seman-
tics searches truth through a process of falsification and verif-
ication. The truth of an expression is determined through a set
of domain-dependent rules which define a "game" between two
agents: one agent is trying to validate the expression, the other
one is trying to refute it. The expression is true if the truth
agent wins. Unlike Dummett's verificationist semantics,
Hintikka's is still a "truth-conditional" semantics.
The existence of a winning strategy for either player can be
expressed in the form of a higher-order sentence. This sentence
asserts the existence of the relevant Skolem functions. Game-
theoretical semantics is therefore a translation of first-order
languages into higher-order languages. Game-theoretical semantics
can be easily extended to intensional logic as a successive step
to possible-world semantics. The transition to natural languages
is performed by substituting proper names for entire quantifier
phrases. In natural languages the application of game rules is
governed by second-order principles.
Hintikka Jaakko: LOGIC OF EPISTEMOLOGY (Kluwer Academics, 1989)
A collection of articles on the (limitations of) semantics of
possible worlds and epistemic logic (logic of knowledge).
Hintikka Jaakko & Sandu Gabriel: ON THE METHODOLOGY OF LINGUIS-
TICS (Blackwell, 1990)
Hintikka presents a case study for his "game-theoretic semantics"
by applying it to the treatment of coreference.
Hintikka Jaakko: ASPECTS OF METAPHOR (Kluwer Academics, 1994)
A collection of papers on metaphor, including Bipin Indurkhya's
argument for an interaction theory of cognition and metaphor,
Noel Carroll's presentation of visual metaphors and Eric
Steinhart's model for generating metaphors in the context of
semantic fields.
Hinton Geoffrey & Anderson James: PARALLEL MODELS OF ASSOCIATIVE
MEMORY (Lawrence Erlbaum, 1989)
A selection of readings on parallel associative memory.
D. Willshaw's "Holography, associative memory and inductive gen-
eralization" notes similarities between neural networks and holo-
grams (such as information is not localized but spread over the
entire system).
Hirst William: MAKING OF COGNITIVE SCIENCE (Cambridge, 1988)
A collection of essays in honor of George Miller.
Hobbs Jerry & Moore Robert: FORMAL THEORIES OF THE COMMONSENSE
WORLD (Ablex Publishing, 1985)
A collection of seminal papers on commonsense reasoning, includ-
ing the official version of Pat Hayes' "The naive physics mani-
festo".
Pat Hayes' "The naive physics manifesto" defines "measure space"
for each quantity (length, weight, date, temperature) as a space
in which an ordering relationship holds. Measurement spaces are
usually conceived as discrete spaces, even if the quantities they
measure are in theory continous. In common use things like birth
dates, temperatures, distances, heights and weights are always
rounded. Unlike McCarthy's situations, Hayes' "histories" (con-
nected pieces of space-time) have a restricted spatial extent,
thereby avoiding some of the inconveniences of situations.
Hayes' logistic approach was very influential in formalizing and
axiomatizing common sense knowledge.
The elementary unit of measure for common sense is not the point,
but the interval. Which interval makes sense depends on the
domain: history is satisfied with years (and sometimes centu-
ries), but birth dates require the day and track and fields races
need tenths of seconds.
The relationships between intervals differ from relationships
between points. Two intervals can partially overlap. An interval
can be open or closed. Points require Physics' differential
equations, but intervals can be handled with a logic of time that
deals with their ordering relationship.
The book includes DeKleer's "A qualitative physics based on con-
fluences", Robert Moore's "A formal theory of knowledge and
action" and James Allen's "A model of naive temporal reasoning".
Allen's representation of time is based on intervals, not
instants. Intervals may be related in several ways: one being
before, after or equal to another.
Hobson J. Allan: THE DREAMING BRAIN (Basic, 1989)
From the five cardinal features of dreams (intense emotion,
illogical content, sensory impression, uncritical acceptance,
difficulty of recalling) and their similarities to mental ill-
ness, Hobson derives a theory of dreams as a theory of mental
illness. Hobson thinks that dreams need not be interpreted: their
meaning is transparent.
Hobson builds a model of the brain-mind which specifies which
brain cells and molecules trigger REM sleep and dreaming and the
dynamics of their interaction. His "activation-synthesis"
hypothesis (periodic activation of the brain by the brain stem
and synthesis provided by the forebrain) assumes that dreams are
meaningful: the mind makes a synthetic effort to provide meaning
to the signals that are generated internally (during a dream
memory is even "hypermnesic", i.e. is intensified). Wishes are
not the cause of the dreaming process, although, once dreaming
has been started by the brain stem, wishes may be incorporated in
the dream. Dreams are generated by internal signals.
The function of dreams is speculated to be that of deriving cru-
cial action patterns from the genetic program of the individual.
Hobson therefore interprets dreams in the realm of neurophysiol-
ogy.
Hofstadter Douglas: FLUID CONCEPTS AND CREATIVE ANALOGIES
(Basic, 1995)
With this book Hofstadter goes as far as to propose a cognitive
model, or at least refuse existing cognitive models, for the
mind. The book comes with the software that was built to imple-
ment these analogical strategies, Copycat.
Hofstadter Douglas & Dennett Daniel: THE MIND'S I (Bantam, 1982)
A collection of articles from philosophers, mathematicians and
novelists, surrounded by HJofstadter's own reflections on the
themes of mind and consciousness.
Hofstadter Douglas: GODEL ESCHER BACH (Vintage, 1980)
A bold synthesis of mathematics, art and music, and a collection
of intriguing thought experiments with recursion, self-reference,
decision theory, artificial intelligence and genetics presented
in a very elegant and creative manner.
Consciousness could be caused by "strange loops", an interaction
between levels in which the top level and the bottom level influ-
ence each other.
Holland John et al: INDUCTION (MIT Press, 1986)
A study of induction (perceived as "how knowledge is modified
through its use") built around a rule-based framework. Induction
is directed by problem-solving activity and based on feedback
about the value of its predictions. Learned categories are iden-
tified by clusters of rules. Induction involves two fundamental
processes: a process to revise parameters of existing rules and a
process to generate new rules. Both processes are guided by
knowledge about the domain.
Classifier systems are message-passing variants of production
systems. A classifier system learns syntactically rules (or
"classifiers") to guide its performance in the environment. A
classifier system consists of three main components: a production
system, a credit system (such as the "bucket brigade") and a
genetic algorithm to generate new rules.
Analogical reasoning is considered as a special case of induc-
tion.
Holland John Henry: ADAPTATION IN NATURAL AND ARTIFICIAL SYSTEMS
(MIT Press, 1992)
Revised edition of the seminal 1975 book that generated the
momentum for the study of complex adaptive systems and genetic
algorithms.
Holland had the intuition that the best way to solve a problem is
to mimick what biological organisms do to solve their problem of
survival: to evolve (through natural selection) and to reproduce
(through genetic recombination).
Genetic algorithms apply recursively a series of biologically-
inspired operators to a population of potential solutions of a
given problem. Each application of operators generates new popu-
lations of solutions which should better and better approximate
the best solution.
What evolves is not the single individual but the population as a
whole.
Genetic algorithms are actually a further refinement of search
methods within problem spaces. Genetic algorithms improve the
search by incorporating the criterion of "competition".
A measure function computes how "fit" an individual is. The
selection process starts from a random population of individual.
For each individual of the population the fitness function pro-
vides a numeric value for how much the solution is far from the
ideal solution. The probability of selection for that individual
is made proportional to its "fitness". On the basis of such fit-
ness values a subset of the population is selected. This subset
is allowed to reproduce itself through biologically-inspired
operators of crossover, mutation and inversion.
Each individual (each point in the space of solutions) is
represented as a string of symbols. Each genetic operators per-
form an operation on the sequence or content of the symbols.
Holland's classifier (which learns new rules to optimize its per-
formance) was the first practical application of genetic algo-
rithms. Its emphasis on competition and coopertation, on feedback
and reinforcement, rather than on pre-programmed rules, set it
apart from knowledge-based models of intelligence.
Holland John: HIDDEN ORDER (Addison Wesley, 1995)
Holland focuses on "complex adaptive systems". Such systems are
governed by principles of anticipation and feedback. Based on a
model of the world, an adaptive system anticipates what is going
to happen. Models are improved based on feedback from the
environment.
Complex adaptive system are ubiquitous in nature. They include
brains, ecosystems and even economies. They share a number of
features: each of these systems is a network of agents acting in
parallel and interacting; behavior of the system arises from
cooperation and competitiong among its agents; each of these sys-
tems has many levels of organization, with agents at each level
serving as building blocks for agents at a higher level; such
systems are capable of rearranging their structure based on their
experience; they are capable of anticipating the future by means
of innate models of the world; new opportunities for new types of
agents are continously beeing created within the system.
All complex adaptive systems share four properties (aggregation,
nonlinearity, flowing, diversity) and three mechanisms (categori-
zation by tagging, anticipation through internal models, decompo-
sition in building blocks).
Holland also reviews his own framework for representing adaptive
agents, consisting of a performance system (to describe the
system's skills), a credit-assignment algorithm (to reward the
fittest rules) and a rule-discovery algorithm (to generate plau-
sible hypotheses). His new visual model is called ECHO, and it
"echoes" the creation of complex structures by natural selection.
ECHO operates on a network of sites, each containing resources
and agents. Each agent's structure is defined in terms of
strings of resources, each string being a chromosome. Each chro-
mosome contains three tags (offense, defense and adhesion), three
conditions (exchange, mating and replication), and a list of
resource transformations. Tags and conditions determine what hap-
pens when two agents interact.
Humphrey Nicholas: CONSCIOUSNESS REGAINED (Oxford Univ Press,
1983)
Humphrey thinks that the function of consciousness is that of
social interaction with other consciousnesses. Consciousness
gives every human a priviliged picture of her own self as a model
for what it is like to be another human.
Consciousness provides humans with an explanatory model of their
own behavior. Psychological skills are a biologically adaptive
trait in human beings: the best psychologists are the best sur-
vivors. The best psychologists are those who have the widest
range of personal experience.
Humphrey Nicholas: A HISTORY OF THE MIND (Simon & Schuster,
1993)
A study of the evolution of consciousness from simple matter to
thought, emotions and self-consciousness.
Humphrey claims that to be conscious is to feel sensations, as
opposed to perceptions. Sensations are to be found at the boun-
dary between the organism and the world and at the boundary of
past and future. One "senses" a circle of light hitting the
retina; one "perceives" the sun in the sky. One can have sensa-
tions about perceptions and perceptions about sensations. Animals
have developed two ways of representing the interaction between
the body and the world: affect-laden sensations and affect-
neutral perceptions.
Sensation and perception are separate and parallel forms of
representation. Consciousness is about sensation. Humphrey
develops a theory of sensations, feelings and actions. The last
stage of the evolutionary journey is a "sensory reverberating
feedback loop" within the brain. Then consciousness arises.
Hutchinson George Evelyn: AN INTRODUCTION TO POPULATION ECOLOGY
(Yale University Press, 1978)
Hutchinson reviews the field of population dynamics, introduces
formal definitions for quantities such as "ecological niche" ("an
N-dimensional hypervolume within which environmental conditions
at every point permit an organism to live") and derives nonlinear
analyses of populations.
The whole theory is based on two postulates: the principle of
abiogenesis (every living organism has originated from at least
one parent of like kind to itself, "omne vivum ex vivo"); and the
postulate of upper limit (there is an upper limit to the number
of beings that can utilize a given finite space). They are both
reflected in Verhulst's "logistic", a mathematical model for a
continously growing population with an upper limit. There exist a
number of variants of the original logistic, mainly to take into
account factors such as competition and coexistence.
Any sulf-sustaining biological community must include on popula-
tion of photosynthetic plants at its lower level. Herbivores feed
on this level and form a new level, on which primary carnivores
feed and form a new level, on which secondary carnivores feed,
etc. Each level is smaller (not only in number but also in
biomass) than the lower one, thereby originating a pyramidal
structure.