A DAY IN THE LIFE OF A MEME

Liane Gabora
Center for the Study of the Evolution and Origin of Life
University of California, Los Angeles
Los Angeles, CA 90095-1567
liane@cs.ucla.edu


To appear in The Nature, Representation, and Evolution of Concepts
(ed. Philip van Loocke) 1997, Routledge Press.  Do not quote without
permission.  Copyright Liane Gabora, 1997.


1. CONCEPTUAL CHANGE: A SECOND FORM OF EVOLUTION

Knowing everything there is to know about how concepts or ideas can
get represented in the mind of an individual would not take us far
toward an account of why individuals understand the world the way they
do, or explain why their understanding differs from that of their
ancestors.  While some ideas fade into obscurity, others rise to
prominence and spread horizontally through society, and vertically
from one generation to another, getting progressively refined and
embellished along the way.  Ideas, like the strands of DNA that encode
instructions for building and maintaining living organisms, seem to
undergo a process analogous to biological evolution. Understanding
this process is a vital part of unraveling our cognitive make-up.
	There has been a slow but steady effort to map the concept of
evolution onto the dynamics of culture.  Popper [1963] and Campbell
[1987] alerted us to the evolutionary flavor of epistemology.  Dawkins
[1976] introduced the notion of a meme--a replicator of cultural
information analogous to the gene.  In his words: "Just as genes
propagate themselves in the gene pool by leaping from body to body via
sperm or eggs, so do memes propagate themselves in the meme pool by
leaping from brain to brain."  Others have drawn from mathematical
models of population genetics and epidemiology to model the spread of
ideas [Cavalli-Sforza & Feldman 1981; Lumsden & Wilson 1981; Schuster
& Sigmund 1983; Boyd & Richerson 1985; Hofbauer & Sigmund 1988].
	These works point toward the possibility that memetics
constitutes a second form of evolution, distinct from yet intertwined
with biological evolution, with the potential to provide the kind of
overarching framework for the social and cognitive sciences that the
first form provides for the biological sciences.  However thus far
memetics has not lived up to this potential, a situation that seems
unfortunate given the success of the biological precedent.  Although
much was known about living things before Darwin, his theory of how
life evolves through natural selection united previously disparate
phenomena and paved the way for further biological inquiry.
	Some believe that looking to biological evolution to gain
insight into cultural evolution is a waste of time.  As Gould [1991]
put it: "Biological evolution is a bad analog for cultural
change... biological evolution is powered by natural selection,
cultural evolution by a different set of principles that I understand
but dimly."  However at an abstract level of analysis they amount to
the same thing: exploration and transformation of an information space
through variation, selection, and replication.  Both present the
question of what underlying mechanisms could &launch a
self-perpetuating adaptive process.  Thus the possibility that the two
have enough in common that the former can pave the way for the latter
seems at least worth pursuing.
	Skeptics may wonder how we can develop a theory of cultural
evolution before we understand how memes are instantiated in the
brain.  This situation has a precedent: Darwin came up with the theory
of biological evolution through natural selection before the discovery
of genes.  It turned out that genes are laid out in a fairly
straightforward way in physical space, which does not appear to be the
case with memes. This does not mean they can't evolve, so long as
there is a way of retrieving the components of a meme so they can work
together as a unit.  We may not yet know all the physiological details
of how the information manifested in, say, a handshake between two
individuals--with its unique arrangement of contact points, applied
forces, and trajectory--can be traced back to these individuals'
mental representations of handshakes, each other, and the situation
they are in.  But every day more pieces of the puzzle fall into place,
and we will proceed with the confidence that the rest of the puzzle
pieces exist and can be found.
	This chapter outlines a theory of how memes evolve, and
illustrates how the memetic perspective provides not only not only a
foundation for research into the dynamics of concepts and artifacts at
the societal level, but a synthetic framework for understanding how
representations are generated, organized, stored, retrieved, and
expressed at the level of the individual.  It also sketches a
tentative theory of how an infant develops a sustained train of
thought and thus becomes a cog in the meme-evolving machinery.
Implications of this theory pertaining to the mechanisms underlying
creativity, and why it is virtually nonexistent in other species, are
explored.  It concludes with discussion of how a cultural-evolution
perspective can shape and inspire research in the cognitive and social
sciences.

1.1 COMPONENTS OF AN EVOLUTIONARY SYSTEM

In order for evolution to happen there must be: 

1. A &pattern &of &information (a state within a space of possible
states).

2. A way to generate &variations of the pattern (explore or transform
the space).

3. A rationale for &selecting variations that are adaptive-- tend to
give better performance than their predecessors in the context of some
problem or set of constraints (a fitness landscape applied to the
space).

 4. A way of &replicating and &transmitting (or amplifying, as
molecular biologists refer to it) the selected variations.

In biological evolution, the evolving patterns of information are
genes encoded as sequences of nucleotides.  Variations arise through
mutation and recombination, and natural selection weeds out those that
are maladaptive.  Replication takes place at the level of the
genotype.  In cultural evolution, the evolving patterns of information
are memes--mental representations of ideas, behaviors, or other
theoretical or imagined constructs, perhaps encoded as patterns of
neuron activation.  Variation arises when representations are
combined, transformed, or reorganized.  Replication is phenotypically
mediated; it occurs when representations are transformed into action
or language, transmitted through processes such as imitation, and
reproduced, more or less, in another brain.  Incorporation of these
new information patterns into the society alters the selective
pressures and constraints exerted by the social environment, which in
turn leads to the generation of yet more patterns.  Thus memes, like
DNA, comprise a self-sustained system for the relentless exploration
and transformation of a space of possible patterns.

1.2 COGNITION IS NOT JUST A PREDICTABLE EXTENSION OF BIOLOGICAL
EVOLUTION

The line of reasoning presented here can be succinctly conveyed in
terms of &information, which is related to the number of differences
required to specify the state of a system [Shannon & Weaver 1963;
Bateson 1972].  States have not only a structure of difference
relations between them, but also a combinatorial structure--each state
can itself comprise an information space--so that complex information
can be built up from simple information.  We can rate each state in a
space of possible states against some performance measure or fitness
criterion and the result is referred to as a &fitness l&andscape, and
we can move from one state to another by way of a &computation.  The
world can be viewed as a vast network of computations wherein
information is created, transformed, and destroyed.  This information
often exhibits &pattern, or statistical regularity that can be
expressed mathematically.
	After seeing many shadows cast by the same object we can
develop an internal model of what that object looks like without
having seen it.  If more than one object is casting shadows we can
learn to tell which object is casting any particular shadow.
Similarly, by viewing every pattern we encounter as a shadow or
footprint of one or more broad causal principles [note 1], we can gain
insight into the causal principles that manifest pattern.
	If you were to go back to some time during the first billion
years of Earth's history, the only causal principle you would need to
invoke to explain pattern in the information present (with the
exception of yourself) would be the &physical &constraints &and
&self-organizing &properties &of &matter.
	If you were to go back to some time &after the origin of life,
approximately three billion years ago, this would no longer be the
case.  Not that life doesn't exhibit the properties of matter.  But it
would be virtually impossible for, say, a giraffe to appear in an
information space not acted upon by natural selection.  Another causal
principle--&biological &evolution--would have to be invoked from this
point on.
	Today the Earth is embedded with artifacts like computer
networks and circuses that cannot be accounted for by appeal to either
the properties of matter or biological evolution. That is, biological
evolution does not provide us with adequate explanatory power to
account for the existence of computers any more than the properties of
matter can explain the existence of giraffes.  Computers are
manifestations of yet another causal principle: &the &evolution &of
&culture.
	Thus pattern in the structure and dynamics of information we
encounter in the everyday world can be traced to three broad causal
principlesQthe physical constraints and self-organizing properties of
matter, biological evolution, and cultural evolution.  This
classification scheme, like all classification schemes, is somewhat
arbitrary.  There may be subclasses of these principles that deserve
to be considered principles unto themselves [note 2], or one could
argue that evolution is a self-organizing property of matter, albeit a
spectacular one [note 3].  The point is: culture is the only process
that has arisen since the origin of life that relentlessly exploits
the combinatorial potential of information.  Despite the fact that
culture is grounded in biology (like biology is grounded in the
physical constraints and self-organizing properties of matter), the
probability of computers arising spontaneously in an information space
not acted upon by cultural evolution (like the probability of giraffes
arising spontaneously in an information space not acted upon by
biological evolution) is vanishingly small.  Thus it is inappropriate
to dismiss culture as a predictable extension of biological evolution.
It is different from anything else biology has produced.
	Since the machinery that renders cultural evolution--the
human brain--is a product of biological evolution, much of what is
'out there' can not be cleanly traced to a biological or cultural
origin.  We will discuss how biology constrains culture through the
preferential spread of memes that satisfy biologically-derived needs.
It goes the other way too; culture not only affects biological fitness
through its effect on behavior (a phenomenon known as the Baldwin
Effect) but it dramatically modifies the biological world.  Some of
the ways in which biological information gets tainted with cultural
information seem relatively inconsequential, such as the trimming of
hedges, whereas others, such as dog-breeding, have a long-lasting
effect.  (In fact, one could view dogs as the consequence of a memetic
trajectory that was launched by the need to protect property.)
Nevertheless, much as it is not imperative to address the role of
physical constraints like gravity in studies of, say, embryonic
development or squirrel foraging behavior, much can be said about
culture without addressing the role of biological constraints.

1.3 TAKING THE MEME-EVOLVING APPARATUS SERIOUSLY

The most salient shortcoming of existing models of cultural evolution,
and quite possibly the reason they have had relatively little impact,
is that although they tell us much about the transmission or
characteristics of 'catchy' memes, they fail to adequately address the
issue of how novelty is generated.  To be of significant theoretical
or predictive value, memetics must give serious consideration to the
processes by which experience in the world turns into new memes in our
brains, and address how memetic novelty is expanded further through
creative processes.
	Studies of creativity, on the other hand, suffer from the
opposite problem.  They tend to focus on the &individual, obscuring
the fact that creativity is a collective affair [e.g. Lenat 1974;
Schank & Leake 1989; Mitchell 1993; Boden 1991].  The ideas and
inventions an individual produces build on the ideas and inventions of
others.  Which memes spread and which ones die off reflects the
dynamics of the entire society of individuals hosting them.
	Although at a sufficiently abstract level the notion that
culture evolves is obvious, we lack a theoretical framework that
bridges transmission studies with studies of creativity, and spells
out explicitly how the concept of evolution maps on to the case of
culture.  To accomplish this we need more than a quick and dirty list
of what makes a meme catchy; we have to consider the extent to which
it resonates with and enriches the complex web of assumptions,
beliefs, motives, and attitudes of its host--no simple matter.  In
short, we have to take the meme-evolving apparatus very seriously.

2. MEME: INFORMATION UNIT THAT EVOLVES THROUGH CULTURE

Because the concept of the individual is so germane to the human
experience we have an anthropocentric tendency to assume that the
individual is the appropriate basic unit of analysis.  This tendency
is probably exacerbated by the fact that in biology the individual is
the object of the relevant evolutionary process (or more specifically
the phenotypic expression of the information undergoing evolution).
The memetic approach involves relinquishing our focus on the
individual, and concentrating instead on the meme as the object of a
second evolutionary process that makes cognition possible.  This
perspective can feel unnatural and disorienting but it discloses
population-level phenomena that would otherwise go unnoticed because
they are not readily detected through introspection.

2.1 DISTINCTION BETWEEN A MEME AND ITS PHENOTYPIC IMPLEMENTATION

Durham [1991] defines a meme as "any kind, amount, and configuration
of information in culture that shows both variation and coherent
transmission."  Problems with this definition arise because it does
not distinguish between cultural information as mental representation
and cultural information as implemented behavior or artifact.
	The genotype-phenotype distinction is useful here.  The
cultural analog of a genotype is the mental representation of a meme,
and the analog of a phenotype is its &implementation, or the form it
takes if it gets expressed or communicated, typically as action or
vocalization.  Implementation transforms a meme, incorporating
syntactic features characteristic of the channel through which it is
conveyed [Brooks 1986].  Thus, for example, a dance step looks
different with each individual who performs it.

2.2 INTRA-MEME STRUCTURE: CORE, ENABLER, AND HITCHHIKER FEATURES

Biologists use the term &allele to capture the notion of alternative
heritable versions of a gene, and Durham [1991] accordingly adopted
the term TallomemeU to refer to alternative versions of a meme.  This
basic concept was tailored to meet the constraints of biology; we all
have the same number of genes, and two alleles of each gene (one from
each parent).  The cultural analog may be too clumsy to capture the
subtle relationships between memes.  Memes often appear to be stored
in a distributed, network-like fashion, connected through webs of
association [Hebb 1949; Quillian 1968; Pribram 1974]; there is not
necessarily a definitive rationale for saying where one stops and
another begins, in semantic space let alone physical space.  For
example would we consider 'My mother looks good in blue' and 'My
mother looks good under a blue umbrella' to be allomemes of the same
meme, or different memes?
	This kind of difficulty is circumvented by avoiding the notion
of alternate versions altogether and using the term &feature to refer
to a component of a meme.  Thus related memes share features.  In this
paper, feature can refer to a component with any degree of granularity
below that of the meme in question; thus the scope of what might be
considered a feature could range from an entire array of visual
information depicting every perceived quality of a particular umbrella
(such as might occur early on in perception) to one bit of information
indicating the presence or absence of an umbrella (such as might occur
at an advanced stage of cognitive processing).  There can be nonlinear
(epistatic) relations amongst the features of a meme, as well as
between a meme and its implementation.
	A first step toward a science of memetics is to decompose
memes into features or feature schemata according to how they relate
to fitness.  Here we will distinguish the following categories: (1)
&core &features, that contribute directly to the fitness of a meme,
(2) &enabler &features, that enable or facilitate the implementation
of core features, and (3) &hitchhiker &features, which exist in the
meme due to arbitrary or accidental historical associations with
features of the first two kinds.  Core features tend to convey
semantic information, and enabler features syntactic information,
though one can think of situations in which some semantic information
serves simply to facilitate expression of other semantic information
i.e. functions as an enabler.  The first two categories are vaguely
analogous to the categorization of genes as structural or regulatory,
and the last category is inspired by the phenomenon of genetic
hitchhiking [Kojima & Schaeffer 1967; Maynard Smith & Haigh 1974;
Kaplan, Hudson, & Langley 1989].  The closer together genes are on a
chromosome the less likely they will be separated by crossover, so the
more tightly linked they are said to be.  Hitchhiker alleles confer no
fitness advantage, but endure because they are linked to alleles that
are important for survival.  In both genetic hitchhiking and its
cultural analog there is indirect selection for useless (or even
detrimental) patterns through their association with beneficial ones.
The concept of hitchhiking is closely related to that of exaptation--
the evolution of organs or traits not evolved through natural
selection for their current use [Gould 1982].

2.3 MEME AS AN EXPERIENCE ENCODED IN THE FOCUS 

In his book &Thought &Contagion, Aaron Lynch claims "Much as computer
viruses make up a small fraction of computer software, so too do
pernicious thought contagions form a small part of our mental
software."  But is this conservative approach necessary, or even
sensible?  Where do you draw the line between potentially
culturally-transmittable memes and the rest our "mental software"?
The haiku-writer, for example, seems able to draw upon any sort of
thought, memory, or imaginary construct.  
	Here we will explore the opposite approach to that taken by
Lynch; we bite the bullet and consider anything that can be the
subject of an instant of experience, to be a meme.  The category
TmemeU now includes not only obviously transmittable ideas like 'Be
good or you will go to Hell', but everything from a particular
experience of vibrant redness, to a realization of a shorter route
to work, to a feeling of elation associated with a friendUs facial
expression or posture.  This may strike some readers as outrageous,
but it doesn't really make things as unwieldy as it might seem to at
first glance.  For the price of this added complication we gain a
bridge that connects memetics with phenomena like perception, body
language, planning, deductive reasoning, emotion, categorization,
analogy... the stuff of the social and cognitive sciences.  It may be
our only viable direction.  A theory of mind that can account for myth
and freestyle dance, not to mention rapid personality assessment, is
not easily achieved.
	Our concept of meme is made more manageable by 
invoking Kanerva's [1988] notion of the &focus--that part of 
the mind in which sensation (either external or internal e.g. 
hunger) and stored memory interact to produce a stream of 
experience.  The states of the neurons that comprise the focus 
determine the content and experiential quality of an 
individual's awareness.  One can think of a meme as a pattern 
of information that is or has been encoded in an individual's 
focus.  It can be subjectively experienced as a sensation, idea, 
attitude, emotion, or combination of these, and it can direct 
implementation by the motor apparatus.  

3. THE MEME-EVOLVING APPARATUS: CONCEPTUAL NETWORKS
  
Memes are woven together in an internal model of the world 
or &worldview, which their host calls upon to figure out 
what to do whenever a situation is too complicated for its 
hardwired instincts.  The worldview orchestrates behavior 
such that a meme gets implemented right when it is likely to 
be useful, and that increases the probability that other hosts 
will consider it worthy of replication.  This section briefly 
discusses some ideas about how the worldview is structured.
	 
3.1 REPRESENTATION OF LOGICAL RELATIONSHIPS 

A worldview with enough internal consistency to put together
strategies and solve problems must have a way of representing not only
memes themselves but the logical relationships amongst them.
McCulloch and Pitts [1943] showed that networks made of neuron-like
components that perform the logical operations AND, OR, and NOT are
theoretically capable of computing any Turing machine-computable
function.  In connectionist-type systems, logical relations are
represented implicitly as constraints on the possible states of a
system, and computation proceeds through settling into a solution that
satisfies many constraints, rather than then explicitly calculating a
solution.  This is accomplished through modification of the
associative strengths amongst the components of the system, and the
process is referred to as relaxation or simulated annealing.

3.2 THE BABY AND THE BATHWATER: CONCEPTUAL LINKAGE DISEQUILIBRIUM

Arguments against a theory of cultural evolution generally consist of
a series of statements as to how the cultural situation differs from
that of biology [e.g. Gould 1991; Thagard 1980].  These arguments,
however, do not constitute a viable reason to discard the idea that
culture is an evolutionary process.  Imagine that 100 years before
Darwin proposed the theory of biological evolution through natural
selection, another scientist had discovered another system whereby
information patterns evolved, say in a test tube.  Given this
scenario, would it have made sense for Darwin to dismiss biological
evolution simply because it works differently from the test tube form?
This would obviously have been foolish.  It would have robbed humanity
of not only the unifying power of a theory of biological evolution,
but the opportunity to use knowledge of how evolution works under one
set of constraints and affordances as a scaffold to direct the study
of how it works under different constraints and affordances.  But time
and again it is argued that a theory of cultural evolution is doomed
simply because it would have to work through different mechanisms from
those of biological evolution.
	Ironically this situation in itself provides us with a nice
example of how knowledge of evolution acquired in the realm of biology
can help unravel analogous situations in the realm of culture.  The
biasing effect of historical association is an important theme in
population genetics.  Alleles of linked genes, such as the those that
code for red hair and freckles, continue to co-occur more often than
chance even after individuals in the lineage from which these alleles
originated begin mating randomly with individuals from other lineages
that did not have these alleles.  One can theoretically measure the
number of generations necessary for these genes to achieve a state of
random association or &linkage &equilibrium, and this process can be
modeled computationally.  Similarly, people often have difficulty
applying an idea or problem-solving technique to situations other than
the one in which it was originally encountered, and conversely,
exposure to one problem-solving technique interferes with the ability
to solve a problem using another technique [e.g. Luchins 1942].
Psychologists refer to this as &mental &set (though it is more
commonly known as 'throwing the baby out with the bathwater'). We
could view mental set as a state of &conceptual &linkage
&disequilibrium.  In the present example, conceptual linkage
disequilibrium hinders our ability to abstract the basic concept of
evolution from its biological manifestation so that it can be applied
with ease to the case of culture.  One could argue that it would make
sense for cultural evolution to be the default form of evolution in
disciplines outside of biology, much as in tropical climates the
default form of skiing is water-skiing rather than snow-skiing.
	Conceptual linkage equilibrium is achieved when all instances
of hitchhiking have been obliterated.  In an influential paper on the
relationship between DNA polymorphism and recombination rates, Begun
and Aquadro [1992] suggested that genetic hitchhiking may have
significant evolutionary impact:
	
	This correlation suggests that levels of neutral variation 
	in many of the gene regions for which variation has been 
	measured have been reduced by one or more hitchhiking 
	events.  Provided that a new selectively favored mutation 
	goes to fixation before another advantageous mutation 
	arises close to it, each fixation will be surrounded by a 
	'window' of reduced polymorphism, the relative size of 
	which is proportional to the rate of recombination for that 
	region of the genome. 
  
The general idea here translates nicely to cognition: if a meme goes
to fixation in a society due to selective advantage conferred by one
or more core features, its enabler and hitchhiker features will also
exhibit reduced polymorphism, and the size of the window will vary
with the extent to which hitchhiker features are conceptually bound to
that meme.  (For example, the basic concept of a typewriter/computer
keyboard was historically enabled by the QWERTY keyboard design, which
has now gone to fixation despite poor performance relative to other
possible keyboard designs.)

3.3 INDIVIDUAL VERSUS SOCIETAL WORLDVIEWS 

Our worldviews overlap to the extent that similar experiences and
genetic make-ups cause our brains to store the same memes and styles
of computation.  But they don't overlap perfectly.  Each host's train
of thought traces out a unique trajectory through conceptual space.
It can be useful to think in terms of not only the worldview of an
individual, but also the worldview of a group or even human society at
large, wherein all frontiers of human endeavor are incorporated.

4. SELECTION AND MEME FITNESS

The next few sections examine in some detail how each of the three
phases of evolution--selection, variation, and transmission--map onto
the case of culture.  Though these phases are discussed one at a time,
it is worthwhile to keep in mind that in culture they are less
spatiotemporally distinct than in biology.  Selection can be coupled
to either the generation of variation, or replication, or all three
can occur simultaneously (for example when paraphrasing).

4.1 MEMES RELY ON BRAINS TO SELECT, VARY AND REPLICATE THEM

Von Neumann [1966] postulated that any self-replicating system
consists of two parts: (1) &uninterpreted &information--a
self-description that is passively copied to offspring, and (2)
&interpreted &informationQinstructions for how to construct offspring.
This turned out to be true of the genetic code; there are genes that
provide instructions to the body for how to sustain itself, and genes
that provide instructions for how, with the help of someone of the
opposite sex, to create a child. But unlike genes, memes do not come
prepackaged with instructions for their reproduction.  They rely on
the pattern-evolving machinery of their hosts' brains to create,
select, and replicate them.  Since we preferentially spread ideas that
satisfy needs, our needs define viable niches for memes to evolve
toward.  As infants we might cry and kick no matter what need is most
pressing, but as children we acquire and continually refine a
repertoire of memes that, when implemented, satisfy various needs.  We
learn that reaching into the cookie jar satisfies one need, shouting
ThelpU satisfies another, et cetera.  Our memes, and the behavior they
elicit, slide into need-defined attractors (regions of stability) in
the memetic fitness landscape.
	The fact that memes are not independently self-replicating
does not prevent them from achieving reproductive success.  In fact it
may ironically work in their favor, because the cognitive machinery
they depend upon not only (1) organizes them into a worldview which
orchestrates their implementation in a way that makes their usefulness
apparent, but (2) transforms them to produce 'offspring-memes'.  Both
tendencies--to reorganize memes, and create new ones--stem from the
need to be able to evaluate and carry out plans of action, and seem to
resurface when needs more directly relevant to survival have been met.
We now turn to how these more pressing survival needs affect meme
evolution.
		
4.2 BRAINS SELECT MEMES THAT SATISFY BIOLOGICAL AND CULTURAL NEEDS

Since many of our needs have a biological basis--e.g. the need for
food, shelter, et cetera--meme generation is largely constrained by
our heritage as products of biological evolution.  Thus the topology
of the memetic fitness landscape largely echoes that of the biological
fitness landscape.  In the short term, the biological fitness
landscape, and thus the memetic fitness landscape, fluctuates
continuously as one need is satisfied and others take precedence [Hull
1943; McFarland & Sibly 1975; Gabora & Colgan 1988; Maes 1991].  For
example, after eating, ideas that pertain to finding food are less
likely.  However over the lifetime of an individual the set of
biologically-based needs remains relatively constant.  The trajectory
of survival-motivated thought can be described as a limit cycle
(periodic attractor) that moves through the set of stable memes whose
implementations satisfy the various biological needs.
	Variation-inducing operations (which will be discussed in the
next section) restructure conceptual space and thus affect the memetic
fitness landscape.  Much as the evolution of rabbits created
ecological niches for species that eat them and parasitize them, the
invention of cars created cultural niches for gas stations, seat
belts, and garage door openers.  As one progresses from infanthood to
maturity, and simple needs give way to increasingly complex needs, the
stream of thought acquires the properties of a chaotic or strange
attractor, which can be viewed as the formation of crevices in the
original limit cycle.  The landscape is fractal (i.e. there is
statistical similarity under change of scale) in that the satisfaction
of one need creates other needsQevery crevice when examined closely
reveals more crevices.  This is analogous to the fractal distributions
of species and vegetation patterns described by ecologists [Mandelbrot
1982; Palmer 1992; Scheuring & Riedi 1994].  An endpoint of a cultural
evolution trajectory turns out to be not just a point in
multidimensional space, but a set of points with their own fitness
metric--a micro-landscape in its own right.  So although the memetic
fitness landscape loosely follows the biological fitness landscape,
there are places where it deviates, and this effect undoubtedly
becomes more pronounced throughout an individualUs lifetime.  This
means that the potential for meme diversity is open-ended.  Fueled by
need and constrained by association we carve out trajectories through
meme space, and because the fitness landscape that guides this process
is fractal, every time that landscape steers the production of a new
meme (or even just a slight variant of a preceding meme), the new meme
in turn redefines the landscape, and so on, recursively.  Thus there
is a continuous coevolutionary interplay between meme and landscape,
and this would seem to contribute to the oft-noted rapidity with which
culture evolves.
  
4.3 HARD-WIRED SELECTION

To the extent that the memetic fitness landscape echoes the shape of
the biological fitness landscape, to which we have been adapting since
life began, cultural selection is built right into our architecture.
Our perceptual and cognitive systems are wired up such that they are
primed to focus on and highlight those aspects of external reality
that are relevant to our survival (or were in the past).  The mental
representations we form reflect that bias [e.g. Hubel & Wiesel 1979;
Marr 1982].
	Second, the associative organization of memory both permits
and constrains meme storage and variation- generating operations.
This is discussed in detail extensively elsewhere [e.g. Van Loocke
1991 and this volume] and will not be addressed here.
	
4.4 MALLEABLE FORMS OF SELECTION

In order to create, or even just understand, a new meme, there has to
be a conceptual framework from within which it will make sense, and a
need, or niche, for it.  Therefore, any relevant precursor or
&proto-memes must first be assimilated [Wallas 1926].  This constraint
amounts to a malleable, or plastic, form of selection on new memes.
	When one host exhibits a meme that another observing host
values, the observing host often displays reinforcing body language
and emits words of encouragement.  Likewise, if the meme is
threatening or inconsistent with valued memes, the observing host's
words and behavior tend to be discouraging.  Our need for social
acceptance makes us more likely to exhibit memes that have been
reinforced and less likely to exhibit those that have been
discouraged.  From a meme's-eye perspective this looks like a subtle
strategy whereby a meme acting through one host selectively shapes the
probability of its implementation in other hosts.
	Selection can also occur after a representation has been
internalized but prior to being phenotypically expressed.  For
example, &mentally &simulating what would happen if an idea were
implemented can weed out unworthy ideas [Nersessian 1993].  The
success of mental simulation varies with the accuracy of ones'
internalized model of the world, but it provides at least a
rudimentary form of selection.
	Finally, selection can operate through &biased &transmission;
that is we choose to imitate certain individuals and not others [Boyd
& Richerson 1985].

5. CREATIVITY: WELLSPRING OF CULTURAL VARIATION

5.1 STRATEGY GUIDES TRAJECTORIES THROUGH MEMETIC FITNESS LANDSCAPE

The existence of an open niche does not guarantee that the niche will
ever be found.  In biology the process by which this happens is
largely random.  Though most mutations and recombinations are
detrimental, &so &many variants are generated that it is not necessary
to be clever about how they are generated.  We could say that
biological evolution is a more breadth-first search algorithm than
cultural evolution because it relies primarily on massive parallelism
rather than strategy.
	In culture, on the other hand, variants are generated
strategically.  We could say that cultural evolution is a more
depth-first approach to searching a space of possibilities.  The
trajectory of a stream of thought is constrained by connections
between representations that are similar or spatiotemporally related
[Schank 1983], which increases the probability that an advantageous
variant is found.  For example, when considering the problem of having
to get out of your car every day to open the garage door, you would
not think about doilies or existentialism, but concepts related to the
problemQ electricity, human laziness, and various openers you have
encountered before.  If you were to spend several months on this
problem, ideas that pertain to openers of various kinds would for you
become a highly active region of conceptual space, analogous to the
uncharacteristically high level of activity (and polymorphism) in a
small portion of the human genome known as the major
histocompatibility complex (MHC) which deals with immune response
[Hughes & Nei 1988].

5.2 WORLDVIEW REORGANIZATION: CHUNKING, CATEGORIZING, AND REDESCRIBING

Frequently many memes get integrated into one through a process
referred to in the psychological literature as &chunking [Miller
1956].  Chunking involves forming associations amongst
previously-learned memes and establishing this constellation of
associations as a new meme in long term memory; it is analogous to the
formation of coadapted genes, or &schemata [Holland 1975].
	Whereas chunking generally refers to the binding of
semantically unrelated memes (as in the memorization of an arbitrary
string of numbers), categorization involves the recognition of
semantic relationships.  Categorization and the resultant hierarchical
structure of knowledge is discussed in other chapters and will not
addressed here in any depth, though it is of relevance to point out
that it seems reasonable to expect that the more extensively memes
have been chunked or categorized, the greater the complexity of what
can be held in the focus at once.  Thus what constitutes a meme (and
thus a feature) will differ amongst individuals, and within an
individual over time.
	During creative thought, memes potentially relevant to a
solution would evoke or activate one another, altering or
strategically (though not necessarily consciously) manipulating them,
a process that is said to involve pattern completion, constraint
satisfaction [Rumelhart & McClelland 1986], and the tweaking,
blending, redescription, abstraction, and recoding of representations
[Hofstadter 1985; Holland et al. 1986; Karmiloff-Smith 1986; 1992; Ram
1993; Clark & Thornton 1995].  Neurophysiological evidence suggests
that creating new contexts for representations, that is manipulating
them, involves hippocampal binding or linking [Squire 1992], and
synchronization [Klimesch 1995], of features encoded by distributed
cortical cell assemblies.
	One could argue that recreation is the re-creation of
information patterns in different domains from the ones in which they
were originally encountered, thereby filtering out conceptual
prejudices that reflect nothing more than mechanical constraints or
historical legacies of the original domain.  Play, intellectual
pursuits, and other creative endeavors are then algorithms for
achieving a state of conceptual linkage equilibrium through mental
operations that, like genetic recombination, increase polymorphism by
reducing fixation through hitchhiking.

5.3 SPARSE DISTRIBUTED MEMORY AS A PLATFORM FOR GENERATING NOVELTY

If the worldview is like a patchwork quilt of memes, thought is like
shining a flashlight onto the quilt such that the light is not
restricted to one patch but can overlap several, distilling what is
relevant from each.  In a train of consecutive thoughts we
spontaneously weave external stimuli with concepts or experiences that
are related, sometimes superficially, and other times through abstract
or metaphorical resemblances.  That is, experience has a thread of
continuity.  &Sparse, &distributed &memory [Kanerva 1988], or SDM, is
a mathematical model of the mechanics underlying the storage and
retrieval of memories that nicely captures these aspects of cognition.
	Kanerva draws an analogy between the focus and a combined
address-datum register in a computer; they both contain data and serve
as a pointer to memory, and can both read from and write to memory.
An instant of experience is encoded in the focus by a high-dimensional
vector of difference relations, or bits, that represent the presence
or absence of some feature, and the mathematics generalizes such that
a pattern of bits can represent a value along some dimension.  The
Hamming distance between two memes is the number of bits that differ.
(So the Hamming distance between 11111 and 11100 is two.)  Since each
meme has an antipode (for example, the antipode of 11111 is 00000),
the space of all possible memes can be visualized as a sphere.  The
address of a meme is the information pattern that specifies where the
meme is stored.
	If L is the number of possible features in a meme, the number
of possible memes is 2^L.  Assuming L is large the size this space is
enormous, so the memory is &sparse in that it stores only a small
fraction of the set of all possible memes.  For example, to construct
a SDM with L=1,000, then out of the 2^1,000 possible addresses, a
workable number of them, say 1,000,000, are chosen at random to be
actual storage locations.  The number of memes at Hamming distance k
away from any given meme is equal to the binomial coefficient of L and
k, which is well approximated by a Gaussian or normal curve.  If meme
X is 111...1 and its antipode 000...0, and we consider meme X and its
antipode to be the TpolesU of the hypersphere, then approximately 68%
of the other memes lie within one standard deviation (sqrt[L]) of the
TequatorU region between these two extremes (FIGURE 1).  As we move
through Hamming space away from the equator toward either Meme X or
its antipode, the probability of encountering a meme falls off sharply
by the proportion sqrt[L]/L.  In our example, the median distance from
one location to another is 424 bits, and 99.8% of stored memes lie
between 451 and 549 bits of any given location.

[INSERT FIGURE 2 ABOUT HERE]
                                                                
	A computer reads from memory by simply looking at the address
in the address register and retrieving the item at the location
specified by that address.  The sparseness of the SDM prohibits this
kind of one-to-one correspondence, but it has two tricks up its sleeve
for getting around this problem.
	First, it feigns content addressability, as follows.  The
particular pattern of 1s and 0s that constitutes a meme causes some of
the synapses leading out from the focus to be excited and others
inhibited.  The locations where memes get stored are memory neurons,
and the address of a neuron amounts to the pattern of excitatory and
inhibitory synapses from focus to memory that make that neuron fire.
Activation of a memory neuron causes the meme to get written into it.
Thus there is a systematic relationship between the memes' information
content and the locations they activate.
	Second, since the probability that the ideal address for
storing a meme corresponds to an actual location in memory is
vanishingly small, storage of the meme is &distributed across those
locations whose addresses lie within a sphere (or more accurately,
hypersphere) of possible addresses surrounding the ideal address
(FIGURE 2).  The radius (in Hamming metric) of this sphere is
determined by the neuron activation threshold.  Each location
participates in the storage of many memes.  In this example we assume
that 10,000 memes have been stored in memory.  Each meme is stored in
1,000 (of the 1,000,000 possible) locations, so there are
approximately 10 memes per location.  The storage process works by
updating each of the L counters in each location; to store a 1 the
counter is incremented by 1, and to store a 0 it is decremented by 1.
These nearly one million operations occur in parallel.
                                              
[INSERT FIGURE 2 ABOUT HERE]

	If after a meme, say meme X, is stored, the individual's
attention is directed toward external stimuli, then nothing is
retrieved from memory.  But to the extent that memory contributes to
the next instant of awareness, the storage of X activates retrieval of
not only X itself but all the other memes that have been stored in the
same locations.  The next meme to be encoded in the focus, X', is
found by determining the best match; that is, by averaging the
contributions of all retrieved memes feature-by-feature.  Whereas the
1,000 retrieved copies of X (and memes similar to X) reinforce one
another, the roughly 10,000 other retrieved memes are statistically
likely to cancel one another out, so that XU ends up being similar to
X.  Though X' is a reconstructed blend of many memes it can still be
said to have been retrieved from memory.  X' can now be used to
address the memory, and this process can be reiterated until it
converges on meme Y that satisfies a current need.  This is how the
SDM accomplishes depth-first search.  The closer Y is to X, the faster
the convergence.  In our example, assuming r = 425, if X and Y are
more than 200 bits apart Y is unlikely to be retrieved, but if they
are 170 bits apart Y will be retrieved in about four iterations.
	Keeler [1988] has shown that SDM is a superset of
Hopfield-type and connectionist models of autoassociative or
heteroassociative memory.  The SDM formulation is used here because it
lends itself to an understanding of the mechanics of phenomena we are
interested in.  Because of how the dynamics emerges from the
statistics, rather than from a central executive, it can cope with
creative and seemingly unmechanical cognitive phenomena such as
wordplay or slips of the tongue.  Moreover it is ideally suited to
handle the problem of sequential access, which will become relevant
when we look at how an infant establishes a train of thought.  To
model the recollection of a sequence, meme X is simply used as the
address to write Y, Y as the address to write Z, and so on.  Working
memory can be viewed as the memes that lie within a given Hamming
distance of the meme in the focus such that they are retrievable
within a certain number of iterations.  Categorization could involve
the identification of a pattern, and readdressing memes that contain
this pattern so that their new addresses put them within working
memory reach of one another.  Kanerva shows that the architecture of
common neural components and circuits in the brain are ideally suited
to implement a SDM.
	In SDM, associations between memes are not explicitly
represented as connection strengths but as proximity in
multidimensional space.  However in the end they amount to the same
thing.  The smaller the Hamming distance between two memes, the higher
the probability that they will be retrieved simultaneously and blended
together in the focus (or one after the other in a chain of related
thoughts).  What allows them to be retrieved simultaneously, however,
is that they are either stored in the same neurons or in neurons with
nearby addresses, which in turn reflects the neurons' connectivity.
Thus factors that affect the storage of a meme will also affect
retrieval of that meme; the two processes are intimately connected.

6. MEME REPLICATION AND TRANSMISSION

The memetic approach to cognition is not incompatible with approaches
that stress the role of innate mechanisms [e.g.  Pinker 1995].
Rather, as Lumsden and Wilson [1981] point out, it builds on this
framework, adding that the study of cognition will flounder until we
admit that the role of transmission is equally undeniable.
Transmission links the memetic processing within an individual to not
only memetic processing in other directly-encountered individuals, but
processing in individuals they encounter, and so on.  The ideas and
inventions any one individual produces build on the ideas and
inventions of others.  This phenomenon is known as the &ratchet
&effect, and its impact is demonstrated in the following example.  If
you were suddenly dropped into the Australian desert, you probably
would not survive for long.  However if you were to run into an
aborigine who grew up learning desert survival skills from her family
and community that had been passed on and improved upon for
generations (such as how to find water in obscure places) you might
survive for some time [note 4].

6.1 INTRA-HOST MEME REPLICATION VIA IMPLICIT POINTERS TO MEMORY

We saw how, unlike genetic material, memes do not contain instructions
for how to make copies of themselves; they replicate when their hosts
teach or imitate one another.  The memes in a SDM-like memory,
however, have a self- replication capacity in the following sense.
The pattern of information that constitutes a meme determines which of
the synapses leading out from the focus are excited, and which are
inhibited--it determines how activation flows through the memory
network--which in turn determines the neurons where the meme is stored
and from which the next meme is retrieved.  Thus embedded in the
neural environment that supports their replication, memes act as
implicit pointers to memory.  These pointers prompt the dynamic
reconstruction of the next meme to be subjectively experienced, which
is a variation of (statistically similar to) the one that prompted it.
It is in that sense that they self-replicate.

6.2 TRANSMISSION IS LAMARKIAN AND PHENOTYPICALLY MEDIATED

Internal replication (with variation) makes cultural transmission
Lamarkian--modifications acquired since the acquisition of a meme can
be passed on to others [Dawkins 1976].  The related point that
transmission is phenotypically mediated, as Dennett [1995] points out,
makes a science of memetics less daunting.  It means that, unlike
biologists, we donUt have to fully understand the nature of mental
representation to study transmission.

6.3 ALL STIMULI SCULPT MEMES

While biological needs affect the focus from the inside, environmental
stimuli impact it from the outside.  The information-based orientation
supports a broader conceptualization of the transmission process than
is generally taken.  Transmission often occurs through imitation of
conspecifics [Smith 1977; Bonner 1980; Robert 1990], or guided
instruction [Vygotsky 1978; Tomasello et al. 1993], but not
necessarily.  For example, does it matter whether a child learns to
peel a banana by watching her mother, or a monkey, or a cartoon
character on TV?  What matters is that the child has a mental
representation of how to peel a banana.  All interaction between an
organism and its environment sculpts the worldview, even if just to
increase the organisms' assessment of the probability of encountering
a certain stimulus.  Therefore all stimuli can potentially affect the
interplay of ideas and emotions that evolve through culture.

7. A SCENARIO FOR THE ORIGIN OF COGNITIVE EVOLUTION

We have discussed how memes evolve through selection, variation,
replication and transmission.  We now address the issue of how this
second form of evolution might have begun in the first place.

7.1 THE ORIGIN OF LIFE AND ITS COGNITIVE ANALOG

The origin of life poses the following paradox: how could something as
complex as a self-replicating molecule arise spontaneously?
Traditional attempts to explain this entail the synchronization of a
large number of vastly-improbable events.  Proponents argue that the
improbability of the mechanism they propose does not invalidate it
because it only had to happen once; as soon as there was &one
self-replicating molecule, the rest could be copied from this
template.  However Kauffman [1993] proposes an alternative scenario
that does not entail the synchronization of numerous improbable
events.  He suggests that life arose through the self-organization of
a set of autocatalytic polymers.  When catalytic polymers interact
with one another their average length increases.  As their length
increases, the number of reactions by which polymers can interconvert
increases faster than the number of polymers.  Therefore a set of
interacting molecules under conditions such as are likely to have
existed at the time life began would inevitably reach a critical point
where there is a catalytic pathway to every polymer present.  Jointly
they form a self-reproducing metabolism.
	We now ask: What is the cultural analog to the origin of life?
One could say it is the point in history when organisms acquired the
capacity for social transmission, but as many authors [e.g. Darwin
1871; Plotkin 1988] have pointed out, although transmission is
wide-spread throughout the animal kingdom, no other species has
anything remotely approaching the complexity of human culture.  Donald
[1991] argues convincingly that the bottleneck in cultural evolution
is the capacity for innovation.  Innovation requires more than a kind
of awareness that integrates survival needs with environmental
affordances, and draws upon memory only to interpret stimuli, or
consult a mental map, or recall how some drive was satisfied in the
past.  It requires an ongoing train of creative representational
redescription and counterfactual 'what if...' type thinking.  This
suggests that the cultural analog to the origin of life was the origin
of the first self-perpetuated, potentially-creative stream of thought
in an individualUs brain.
	When an infant has its first experience, there is nothing in
memory to draw upon to contribute to that experience; the first meme
to occupy its focus does not remind it of anything.  Therefore
experience is initially driven only by external or internal stimuli,
not by memory.  Thus the evolution of culture poses a paradox
analogous to that of the origin of the self-replicating moleculeQhow
does an infant develop the capacity for a self-sustained train of
thought that creatively integrates new experiences with previous ones?
Consistent with Kauffman's assertion that the bootstrapping of an
evolutionary process is not an inherently improbable event, the 'it
only had to happen once' argument does not hold water here because the
cultural analog to the origin of life takes place in the brain of
every infant.
    
7.2 ESTABLISHING AN AUTOCATALYTIC SET OF SPARSE, DISTRIBUTED MEMORIES

This section outlines how a SDM-like stream of thought might get
established.  Let us say that the first meme to occupy an infant's
focus and then get stored in memory is a visual experience of its
mother in a blue coat.  The next is the sound of a dog barking.  The
Hamming distance between these memes exceeds the maximum for one meme
to evoke the memory of another, so the barking does not remind the
infant of its mother.  Later the infant sees its mother in a red coat.
This meme evokes or TcatalyzesU the memory of its mother in a blue
coat.  To avoid getting stuck in an endless loop wherein 'mother in
blue coat' then evokes 'mother in red coat' et cetera, it may form the
category 'mother'.  However &that meme does not remind it of anything,
so this stream of thought dies off quickly.
	As the infant accumulates memes, the statistical probability
that a meme in the focus will activate a meme from storage increases,
so the streams of remindings get longer.  Eventually the memory
becomes so densely packed that any meme that comes to occupy the focus
is bound to be close enough in Hamming distance to &some previously-
stored meme(s) to activate a variant of itself.  This marks a phase
transition to a state in which, just as with the origin of life, the
sequential activation of self-similar patterns is self-propelled; the
memes now form an autocatalytic set.  The focus is no longer just a
spot for coordinating stimuli with action, but a forum for the
variation-producing operations that emerge naturally through the
dynamics of iterative retrieval.  The resultant memes evolve along
different trajectories toward different basins of attraction,
TspecializingU in the fulfillment of one need or another.  Those that
satisfy the same need compete until one becomes habitual, while those
that fulfill different needs are able to coexist within the same host.
As with biological speciation, small differences are amplified through
positive feedback leading to transformation of the space of viable
niches for the evolution of information patterns.
	 Note that in this example the 'mother' meme is the infantUs
first category.  A simple way of describing this situation is: if the
'mother in blue coat' meme is represented as 111, and the 'mother in
red coat' meme is represented as 110, the 'mother' meme can be
represented as 11*, where * means either 1 or 0.  It is also the
infant's first derived meme.  That is, it is the first information
pattern to enter the focus not purely by way of external stimuli but
through the necessity of a logical operation on previously-stored
memes--in this case an OR gate--which could be realized in the brain
via adjustment of connection strengths.  The act of categorization
projects the original information space, which had n relevant
dimensions, onto a new space that has n-1 dimensions (for example,
here coat color is no longer relevant).  It effectively makes the
space denser, and increases susceptibility to the autocatalytic state.
On the other hand, creating new memes by &combining previously-stored
memes could interfere with the establishment of a sustained stream of
thought by increasing the dimensionality of the space, thereby
decreasing density.  If indeed cross-category blending disrupts
conceptual autocatalysis, one might expect it to be less evident in
young children than in older children, and this expectation is born
out experimentally [Karmiloff-Smith 1990].
	Note also that the density of memes necessary to reach and
maintain this autocatalytic state will depend on the neuron activation
threshold.  If the threshold is too high (the hypersphere of
potentially activated memes is too small) even very similar memes can
not evoke one another, so a stream of remindings, if it happens at
all, dies off readily.  If the threshold is too low (the hypersphere
too large), then any meme will evoke a multitude of others not
necessarily meaningfully related to it.  Successive patterns in the
focus will have little or no resemblance to one another; the system
may be catalytic but it is not autocatalytic.  The free-association of
the schizophrenic [see Weisberg 1986] seems to correspond to what one
might expect of a system like this.  For memory to produce a steady
stream of meaningfully-related yet potentially creative remindings,
the threshold must fall within a narrow intermediate range.  This is
consistent with LangtonUs [1992] finding that the information-carrying
capacity of a system is maximized when its interconnectedness falls
within a narrow regime between order and chaos.  The situation may
turn out to be slightly more complicated; sustaining a creative train
of thought may involve not only keeping the activation threshold
within a narrowly-prescribed range but dynamically tuning it in
response to the situation at hand.  This is particularly likely if the
memory is not uniformly dense (i.e. clusters of highly-correlated
memes) or if different kinds or stages of thought require different
degrees of conceptual blending.  For example, finding unusual
associations may depend on the preconscious ability to temporarily
increase hypersphere radius.
	Thus we have a plausible scenario for how cultural evolution,
like biological evolution, could have originated in a phase transition
to a self-organized web of catalytic relations between patterns.

8. WHY IS CULTURE UNIQUE TO HUMANS?  A SPECULATIVE ANSWER

Recall that in order for a network of memes to reach an autocatalytic
state, the activation threshold must be calibrated to fall within a
narrow range to achieve a delicate balance between the capacity for
semantic continuity on the one hand and creative association on the
other.  The penalty for having too low a threshold would be very high;
successive thoughts would not necessarily be meaningfully related to
one another, and thinking would be so muddy that survival tasks are
not accomplished.  Too high a threshold, on the other hand, would not
be life-threatening.  The focus would virtually always be impacted
with external stimuli or internal drives such as hunger; memory would
be pretty much reserved for recalling how some goal was accomplished
in the past.  A stream of experience that involved the iterative
reorganization of stored memes would likely die out before it produced
something creative.  This may be the situation present in most brains
on this planet, and the reason that apes are limited to episodic
memory [e.g. Donald 1991].
	The advantages of a sustained train of thought would be
largely lost on animals because they have neither the vocal apparatus
nor the manual dexterity and freedom of upper limbs to &implement
complex ideas.  No matter how brilliant their thoughts were it would
be difficult to do something useful with them.  Moreover, in an
evolutionary line there is individual variation, so the lower the
average activation threshold, the higher the fraction of individuals
for which it is so low that they do not survive.
	These ideas are of course very speculative.  However it seems
reasonable to suggest that apes are not a priori prohibited from
evolving complex cognition, but that there is insufficient
evolutionary pressure to tinker with the activation threshold until it
achieves the requisite delicate balance to sustain a stream of
thought, or to establish and refine the necessary feedback mechanisms
to dynamically tune it to match to the degree of conceptual fluidity
needed at any given instant.  It may be that humans are the only
species for which the benefits of this tinkering process have
outweighed the risks.

9. A COMPUTATIONAL APPROACH

Meme and Variations, or MAV [Gabora 1995] is a computer model of a
society composed of interacting neural network- based agents.  Unlike
other such models that combine biological and cultural evolution
[e.g. Ackley 1994; Spector & Luke 1996] these agents donUt have
genomes, and neither die nor have offspring, but they can invent,
implement, and imitate memes.  MAV successfully evolves patterns of
information through cultural implementations of variation, selection,
and replication, and exhibits phenomena observed in biological
evolution such as: (1) drift--random statistical bias due to sampling
error [Wright 1951; Cavalli-Sforza & Feldman 1981] (2) epistasis
increases time to reach equilibrium, (3) increasing frequency of
variation-generating operations increases diversity, and (4) although
in the absence of variation-generating operations meme evolution does
not occur, increasing variation-generation much beyond the minimum
necessary for evolution causes average fitness to &decrease.  MAV also
addresses issues specific to cultural evolution, such as the effects
of mental simulation, imitation, and strategy.  Perhaps the most
interesting finding it yielded was that although for the society as a
whole the optimal creation-to-imitation ratio was approximately 2:1,
for the agent with the fittest memes, the less it imitated (i.e. the
more computational effort reserved for creation) the better.
 	MAV will hopefully serve as a stepping stone to more advanced
models of memetic evolution.  Of particular interest will be models
that: (1) like Tierra, a model of biological evolution [Ray 1991],
harness the power of evolution to explore and transform an &open-ended
space of possible patterns, but (2) explore the space strategically on
the basis of accumulated knowledge rather than at random5, (3) have
fitness landscapes that emerge through the needs of the agents within
the constraints of their environment, and (4) have agents that must
learn for themselves which memes, when implemented, best satisfy each
of their various needs.
	Mathematical models of culture are too minimal to cope with
the open-ended diversity of culturally-derived information (variation
is generally restricted to trial and error learning or transmission
error) let alone address the numerous intra-individual factors that
undoubtedly have emergent inter-individual consequences, such as how
representations are grounded in experience and how they are stored,
retrieved, and implemented.  Models of individual intelligence and
creativity, on the other hand, lack transmission and replication.
Although this research may not explicitly attempt to address group
processes it typically focuses not on the sorts of simple inferences
and creative acts that a person raised alone in the wild would be
capable of, but on complex acts such as story comprehension, that
might be unlikely to develop in isolation.  With the advent of
massively parallel computers it is becoming increasingly feasible to
place computational models of individual creativity and
problem-solving in a cultural context.  This approach could provide
insight into not only problems pertaining to representation and
culture, but evolution in general, through comparison with biology.
For example, the question of why there is so much redundancy in the
genetic code has generated much discussion which may also apply to the
question of why there are redundant mental maps in the brain; both may
reflect constraints on the nature of an information-evolving code.

10. A MEMETIC PERSPECTIVE ON PSYCHOLOGICAL CONCEPTS

We turn now to how the cultural evolution perspective can shed light
on some aspects of the human experience.

10.1 MENTAL CENSORSHIP AND INDUCTION

Initially an infant is unselective about meme acquisition, since (1)
it doesnUt know much about the world yet, so it has no basis for
choosing, and (2) its parents have lived long enough to reproduce, so
they must be doing something right.  However just as importing foreign
plants can bring ecological disaster, acquisition of a foreign meme
can disrupt the established network of relationships amongst existing
memes.  Therefore the infant develops mental censors that ward off
internalization of potentially disruptive memes.  Censors might also
be erected when a meme is found to be embarrassing or disturbing or
threatening to the self-image [Minsky, 1975]. In a SDM-type
architecture this could be accomplished by increasing the activation
threshold so as to prematurely terminate the relaxation process and
prevent the content of the focus from assimilating with stored memes.
Much as erecting a fence keeps people on one side or the other,
erecting a censor around a meme would warp the statistical probability
that it will partake in the ongoing process of associative recall such
that its host either avoids the censored meme or dwells on it
excessively (FIGURE 3).  This seems to be consistent with our bipolar
attitude toward highly censored subjects such as aggression and
sexuality.
                                              
[INSERT FIGURE 3 ABOUT HERE]

	On the other hand, when the cost of the disruption is
outweighed by the potential benefit accrued by a world model that can
accommodate the new meme, the threshold would be lowered.  Most
thoughts seem to have little effect on our understanding of the world
at large, but once in a while we experience a meme that significantly
modifies our worldview.  The situation is reminiscent of
superconductivity; lowered resistance increases correlation distance,
and thus a perturbation to any one pattern can percolate through the
system and affect even distantly-related patterns.  It would be
interesting to determine experimentally whether the TinductivenessU of
our memes, like other self-organizing systems, exhibits the ubiquitous
inverse power law [Bak, Tang & Weisenfeld 1988].  Just as in a sand
pile perched at the proverbial Tedge of chaosU once in a while a
collision between two grains will lead to another in just the right
chain reaction to generate a large avalanche, occasionally one thought
will trigger a chain reaction of others in a way that reconfigures the
conceptual network.

10.2 THE UNCONSCIOUS

The concept of the unconscious has been influential and useful despite
the obvious incongruity: how is it that we can consciously discuss
something that is unconscious?  What we may be referring to is the
fleeting experience of memes that are dynamically reconstructed as in
a SDM but which do not readily assimilate with other memes and so get
discarded from the focus.  In other words, the need for worldview
consistency prohibits further computational resources from being spent
on trying to integrate what appears to be a nonsensical construction
into the memory.  Of course there is no reason why a meme that is not
immediately integrated into the memory might nevertheless affect the
memory; the very process of determining whether it can be assimilated
or not might itself have effects that infiltrate the system.  This
possibility is supported by the finding that subjectsU behavior can be
affected by priming material of which they have no recall [e.g. Dunbar
& Schunn 1990; Fehrer & Raab 1962].  Subconscious processing of this
sort could, in fact, resculpt the memetic fitness landscape in such a
way that a previously- discarded meme is more readily assimilated the
next time it is encountered.

10.3 CULTURAL MOMENTUM

Despite being derived, directly or indirectly, from human need, memes
do not always promote our survival [Greene 1978; Alexander 1980].  As
Dawkins [1982] points out, "It is true that the relative survival
success of a meme will depend critically on the social and biological
climate in which it finds itself, and this climate will certainly be
influenced by the genetic make-up of the population.  But it will also
depend on the memes that are already numerous in the meme-pool."  Much
like runaway selection in biology, once a meme can replicate with
variation on the basis of some selection criterion, it can evolve out
of the orbit of the need that originated it.  We can't &help but
engage in a stream of thought, spontaneously generating new memes like
'if only such and such had been different...', any more than
biological evolution can help but generate new species.  This
&cultural &momentum could explain why, despite the intuition that
individuals control their streams of thought, creators often express
surprise at the sudden appearance of an idea, and deny active effort
in its immediate creation [Bowers et al 1990; Guilford 1979; Kubose et
al 1980; Wallas 1926].  We seem to control the birth of TourU ideas
only to the extent that we provide a fertile ground for them to be
fruitful and multiplyQby internalizing relevant background knowledge,
identifying new needs, and exposing ourselves to stimuli that help
trigger ideas that fulfill those needs.  (So if you donUt like this
idea, don't blame me.)
	 Spurious basins of attraction sometimes arise in recurrent
neural networks through the compositional interaction of
explicitly-trained attractors [Hopfield 1982].  Cultural momentum may
boil down to a phenomenon of this sort.  Just because the memetic
fitness landscape largely echoes the biological fitness landscape,
that doesnUt mean that behavior elicited by memes in spurious basins
of attraction arising through representational redescription need
always be conducive to survival.  Nevertheless a stream of thought
could be censored before it elicits harmful behavior.  Streams of
thought probably get blocked or sidetracked on a regular basis, not
just by censors, but by minute-to-minute undulations in the
hyperdimensional fitness landscape, that is, change in the relative
urgency of the multitude of survival-related or derived needs
impacting the focus.
	The concept of cultural momentum sheds light on the issue of
free will.  Those who argue for the existence of a central executive
in memory may come to be viewed as the creationists of philosophy and
cognitive science.  Human will can instead be viewed as the emergent
orchestration of needs, stimuli, and retrieved memories impacting the
focus, which is subject to cultural momentum and therefore, in a
sense, beyond our control.

10.4 THE BIRTH OF IDEAS

The biologically-inspired model developed here supports a variant of
the combination theory of creativityQthat new ideas arise through
combinations and transformations of old ones [Boden 1991; Koestler
1964].  The aspect of this theory that does not ring true is that it
neglects the role of emotions.  Here we consider emotions, as well as
ideas, to be encoded as information in memes; some components of a
meme are simply interpreted by parts of the mind that experience them
as emotion, whereas others are interpreted by parts of the mind that
experience them as ideas.  Much research on analogy deals with how the
structure or conceptual skeleton underlying one idea gets abstracted
and applied to another [Gentner 1983; Gick & Holyoak 1983].  We can
expand on this general idea by suggesting that many forms of creative
expression begin with the (unvoiced) question: What would the pattern
of information that encoded the emotion I experienced during this
particular event look like if expressed through the constraints of
that medium?  The existence of inherent limitations on how a pattern
could translate from one domain to another is consistent with the
frequent observation that creativity involves both freedom and
aesthetic constraint. Thus all creativity is directly or indirectly
derived from experience in the world, and since the mathematics
underlying this world, the set of all natural functions, is a small
subset of all &possible functions, the constraints that guide creation
are not arbitrary but objective and familiar; for example the drum
beat of a song might echo a heartbeat, when the rhythm and chord
progression are reminiscent of the sound of someone sobbing we feel
sad, and we hear the wrong note even if we have never heard the song
before.
	It makes sense to expect that a meme or meme complex that has
been censored (for example because it evoked unbearable sadness or
anger) would be vulnerable to being targeted as an area where
worldview cohesion could be increased.  Since at the time the censored
material was experienced it was prohibited from forming associations
to obviously-related memes, it in turn can not be &retrieved through
these expected or straightforward associations.  It can only be
retrieved via Tbackdoor entrancesU, that is associations that reflect
structural congruity at an abstract level.  Thus a musician may come
to habitually funnel patterns encountered in a variety of domains--and
particularly censored material--through modules that filter out
hitchhiker and enabler features, and adapt the core features (or
feature schemata) to the constraints of music.  It is in this
repackaged format that memes originally deemed dangerous can be
integrated into the memory at large without harm, and it is through
this process that the creator establishes a sense of control over
memes that were previously off-limits.
	The account proposed here may seem too simple to explain the
seemingly limitless human potential for creativity, but it may seem
less far-fetched when we consider the variety of species produced by
biological evolution, which operates without the benefit of strategy.
Furthermore, raw materials for the creative process may be acquired in
exceedingly subtle ways.  It is conceivable that you might watch a
stream flow and without your consciously thinking, 'It flows... things
can flow... I could even, in some sense, adopt a more flowing approach
to life', the experience might be reconfiguring your memetic
infrastructure in a way that makes you more easygoing.  I am not
making any claims about the extent to which experiences of this sort
affect us or even whether they occur at all, but rather suggesting
that we not prematurely place a lid on the kinds of processes that
could affect a network of representations and thereby affect creation
and transmission.

10.5 MEMETICS AS MISSING LINK BETWEEN SCIENCE AND SPIRIT

Recall that to implement a meme is to express it, so that it
crystallizes from the world of ideas into words or body language or
objects in the physical world.  Memes fool potential hosts into
wanting or identifying with them, by aligning themselves with memes we
already hold dear (as advertisers are well aware).  The more we value
ourselves in terms of the memes and implemented artifacts we possess
or lack, the more vulnerable we are to ever-more-seductive forms of
persuasion and advertising which tie up time, energy, and resources
that could be applied toward other goals.
	One way to defend oneself against painful or manipulative
memes is to construct what Dennett [1995] refers to as a
&meme-immunological &system; that is, formulate new memes specifically
to deflect &memetic &antigens.  However constructing &memetic
antibodies& of this sort is time-consuming, and like any immunological
response it has to be repeated every time the outside agent evolves a
counter-response.
	Perhaps this explains the purported benefits of transcending
the ego [e.g. Walsh & Vaughan 1993], which can be taken to mean
getting in touch with who we were before our minds were colonized by
memes, through practices such as meditation.  These practices may work
by giving the brain time to anneal material that was never fully
assimilated because of distraction or censorship--mend flaws in the
fabric of the individual's worldview.  Release from the restrictive
power of censors may then enhances the individual's raw 'sense of
self' and produce a feeling of unity or one-ness.
	It is almost impossible to discuss creative insight or
spiritual unfoldment without drawing upon metaphors involving
light. For example: her face lit up, he glowed with enthusiasm, moment
of illumination, reflect on an idea, dim-witted person, and dark
night of the soul, to mention a few.  Even nonverbal communication
uses these metaphors; everyone knows that when a lightbulb appears
above Charlie Brown's head he has had a 'bright' idea.  Experiences of
Tinner lightU have been documented since the dawn of civilization.
Some cultures have a name for it; Eskimo shamen call it qaumaneq,
while Vedanta Hindusists refer to it as Atma.  Lao-tzu advised his
people "Use the light that is within you".  The Gnostics described
flashes of insight or discovery as "light sparks" or &scintillae, and
they conceived of the soul as a radiant sphere that is prone to
contamination, in which case it darkens and becomes more material.
Indra's Net is a Buddhist allegory about a net of threads stretching
horizontally through space and vertically through time, and at every
crossing of the threads is a crystal bead that represents an
individual.  Buckminster Fuller spoke of the soul as a bright sphere
of awareness that moves wherever the individual moves.  Jung suggested
we think of ego-consciousness as "surrounded by a multitude of little
luminosities".
	Since we know the physical world largely through how light of
various frequencies reflects off objects, it is perhaps natural that
we use something analogous to also make sense out of the world of
abstract ideas and emotions.  In diagrams of SDM or neural networks,
when a meme or meme feature becomes active it is generally represented
as lit up.  These diagrams may be derived from the same kind of
intuitive understanding as the mystical experiences mentioned above.
They may turn out to be early prototypes for a portrayal of spiritual
knowledge that is real in the same way as a pie chart of business
expenses represents something real.  We may be on the doorstep of an
era when spiritual experience not only rings true, but makes logical
sense as well.

10.6 MEMETICS AND WOMENUS STUDIES 

One of the things I hope the study of memetics will one day shed light
on is the cumulative effect on womens' self esteem of encountering the
default 'he' or 'his', or phrases like 'man and his world'.  Each time
it happens, it reinforces, however subtly, the thought "I'm not one of
the important ones", or "that doesn't pertain to me so I don't really
count".  The effect of this may be insignificant the first few, or
even hundred times it happens.  But by the time a young girl reaches
adulthood, experiences of this kind probably number in the millions,
and that may be when the effects start to kick in.

11. A MEMEUS-EYE VIEW

In this section we examine a hypothetical and admittedly speculative
scenario for how evolutionary concepts borrowed from biology might
apply to the dynamics of a specific meme.
 
11.1 THE ONTOGENY OF A MEME

One day a classroom bully named Tony put his arm around a 
girl named Memela.  Memela felt threatened by Tony's 
advance.  Her first impulse was to get angry but she censored 
this reaction.  The need for worldview cohesion motivated a 
desire to escape the restrictive power of this censor and find a 
Tbackdoor ventU for her anger.  Eventually she whispered to a 
classmate: "Tony Testosteroni made a pass at me."  Thus 
began the era of Memela's meme.
	Memela's meme can be traced back to a number of precursor or
protomemes, many of which originated in the minds of family and
friends and were subsequently transmitted to Memela.  These protomemes
provided an environmental niche in which the joke could flourish.  But
Memela was more prone to this kind of humor than her siblings.  A sort
of &conceptual &speciation had taken place in their brains, wherein
small initial differences were amplified through positive feedback,
and Memela had become known as the jokester.
	Perhaps Memela was predisposed to produce the joke because her
brain spontaneously exerts a high degree of control over its
activation threshold.  Her ability to find unusual associations by
increasing hypersphere radius, and subsequently refine a train of
thought by decreasing the radius, facilitates word play such as the
establishment of epistatic relationships between semantic and
syntactic features of a meme.  The semantic applicability of the word
testosterone to Tony's aggressive behavior, and the alteration of this
word to make it sound Italian and echo the syntax of (i.e. rhyme with)
the Italian first name, contribute to the humor of Memela's meme.  Its
relation to the highly censored subjects of aggression and sexuality
may also have added to its appeal.

11.2 A DAY IN THE LIFE OF MEMELA'S MEME 

This kind of conceptual epistasis provokes laughter, which draws
attention to a meme.  Memela's meme took full advantage of this.  When
Memela told the joke to one of her classmates the ensuing laughter
attracted a small crowd of other potential hosts, and within an hour
Memela's meme had reached most of the students in the classroom.
Their willingness to invest time acquiring this meme was a smart move;
the joke not only provided amusement, but it proved to be a useful
precursor to the formulation and understanding of subsequent jokes in
this social circle.  Some were direct descendants of Memela's meme,
such as jokes along the lines of "What's for lunch--rigatoni a la
testosteroni?"  Others were more distantly related, such as nicknames
for other classmates that arose because Memela's meme had activated
the general concept of nicknaming.  Across the classroom, ideas that
pertained to nicknaming came to constitute a highly active region of
conceptual space.
	By the time recess ended it had migrated extensively through
the school population.  There were certain subpopulations of
individuals it failed to penetrate, such as social outcasts who were
excluded from much of the memetic exchange.  These individuals' minds
exhibited a cultural version of the Founder Effect [Holgate
1966]--reduced variation in a small population due to genetic drift.
Memela's meme also failed to reach students who engaged in projects
that took them away from the playground at recess.  However since
these individuals had had less opportunity to witness TonyUs behavior,
they had less need to diffuse their fear of him.  Thus even if they
had heard Memela's meme they did not possess the necessary precursor
memes to fully appreciate it; the prerequisite memetic niches were not
in place.  At any rate despite its failure to reach these subgroups,
Memela's meme experienced a high degree of Tmemetic fitnessU.  It
migrated far beyond the classroom in which it was originally
formulated, reconfiguring networks of representations in ways that
affected the subsequent thought and behavior of a number of
individuals.  The telling of this meme and its various incarnations
constituted an act of memetic altruism between like-minded individuals
analogous to the biological altruism that occurs between
genetically-similar individuals.  It played a small role in an ongoing
network of positive reciprocal interactions through which there
emerged a memetically-derived social structure wherein individuals
that regularly generated pleasurable or powerful memes came to be
observed carefully and imitated frequently, while other individuals
were ignored.  Thus the fate of Memela's meme and its descendants
reflected the social and psychological dynamics of an entire society
of interacting individuals.
	Like the other students in the classroom, Tony was affected by
the sound of laughter advertising the presence and amusement value of
Memela's meme.  However it wasn't until someone told him that it was a
joke was about him that he felt willing to do almost anything to hear
it.  Upon hearing the nickname Tony felt ridiculed.  In the case of
Memela's meme a meme-immunological system could be something along the
lines of "That nickname is silly and stupid."

12. CAN MEMETIC EVOLUTION PROVIDE A SYNTHETIC FRAMEWORK FOR THE
COGNITIVE AND SOCIAL SCIENCES?

This chapter gives us a taste of how the concept of evolution may be
able to provide the kind of overarching framework for the humanities
that it provides for the biological sciences.  It sketches out a
theoretical framework for cultural evolution, adopting an approach
analogous to that of the population geneticist in that it emphasizes
meme evolution through social interaction and de-emphasize individual
hosts.  Although the cultural evolution of memes operates through very
different mechanisms from those of biology, culture is the only system
comparable to biology, because it is the only other system to exhibit
the imperative features of evolutionQ adaptive exploration and
transformation of an information space through variation, selection,
and transmission.  All pattern in the information we encounter can be
traced to either (1) the physical constraints and self-organizing
properties of matter, (2) biological evolution, (3) cultural
evolution, or (4) interactions between these causal principles.
	One important difference between the two forms of evolution is
that culture is less random--new patterns have a greater-than-chance
probability of being more fit than their predecessors.  The reason for
this is interesting.  Since memes (unlike genomes) do not come
packaged with instructions for their replication, they must rely on
the pattern-evolving machinery of our brains to do it for them.
Ironically, this state of dependence enhances their proliferative
potential, because the machinery they depend upon constructs and
continually updates mental models of its world--that is, weaves the
memes into a worldview--and uses this worldview to enhance the
assimilation and implementation of the memes and their offspring.
	Cultural evolution presents a puzzle analogous to the origin
of life: the origin of a self-sustained stream of potentially-creative
thought in an infantUs brain.  The idea that life originated with the
self-organization of a set of autocatalytic polymers suggests a
possible mechanism for how this comes about.  Once a threshold density
of assimilated memes is surpassed, any meme that occupies the focus is
close enough in Hamming distance to evoke or 'catalyze' the
spontaneous retrieval or creative reconstruction of a statistically
similar meme, thus the memes form an autocatalytic set.  Note that
this macroscopic account suggests an explanation for only that aspect
of human consciousness that differentiates us from other
'experiencers'; it does not address the mystery of 'raw awareness'
that some say characterizes not only our experience but that of a cow
or a mosquito or even a thermostat [e.g. Chalmers 1996].  Whether or
not this specific theory turns out to be correct, it illustrates how
the analogy to biology can focus our study of culture by providing a
scaffold around which explanatory theories can be built.
	The cultural framework for cognition developed here suggests
tentative explanations for psychological phenomena such as censorship
and creativity, and why they are virtually unique to humans.  It also
suggests explanations for various social phenomena.
	There is good reason to view the mind as a memetic host--a
generator, selector, and replicator of cultural information--and
society as a continually-regenerated constellation of such hosts.
Interactions in the world affect our memetic structure, which in turn
affects future interactions, and these interactions allow cultural
information to evolve, recursively and in parallel in its many hosts.
With each new generation of hosts, the conceptual infrastructure from
which memes are born and grow reflects how memes were encoded and
implemented by the previous generation.  I can not imagine a more
exciting role to play in this process than partaking in the
exploration of memes about memes.

ACKNOWLEDGMENTS
   
I would like to thank David Chalmers, Anthony Francis, Peter Greene,
Pentti Kanerva, Ashwin Ram, Bruce Sawhill, Roger Schank, Patrick
Tufts, and Philip Van Loocke for thoughtful comments on the
manuscript.


REFERENCES 

Alexander, R. D. (1980) Darwinism and human affairs, Pitman.

Ackley, D.  (1994) A Case for Distributed Lamarkian Evolution.  In
Artificial Life III, Addison-Wesley.

Bak, P., Tang, C., & Weisenfeld, K.  (1988) Self-organized
criticality.  Phys. Rev. A 38: 364.

Bateson, G.  (1972) Steps to an ecology of mind, Ballantine Books.

Begun, D.J. & Aquadro, C.F.  (1992) Levels of naturally occurring DNA
polymorphism correlate with recombination rates in D. melanogaster.
Nature, 356, p. 519- 520.

Boden, M. A.  (1991) The creative mind: Myths and mechanisms,
Cambridge University Press.

Bonner, J. T.  (1980) The evolution of culture in animals, Princeton
University Press.

Bowers, K.S., Regehr, G., Balthazard, C. & Parker, K.  (1990)
Intuition in the context of discovery.  Cognitive Psychology,
22:72-110.

Boyd, R. & Richerson, P. J.  (1985) Culture and the evolutionary
process, The University of Chicago Press.

Brooks, V.  (1986) The neural basis of motor control, Oxford
University Press.

Campbell, D. T.  (1987) Evolutionary epistomology.  In Evolutionary
epistomology, rationality, and the sociology of knowledge,
eds. G. Radnitzky and W. W. Bartley, 4 7- 89, Open Court.

Cavalli-Sforza, L. L. & Feldman, M. W.  (1981) Cultural transmission
and evolution: A quantitative approach, Princeton University Press.

Chalmers, D.J.  (1996) The conscious mind: In search of a fundamental
theory, Oxford University Press.

Clark, A. & Thornton, C.  Trading spaces: Computation, representation,
and the limits of uninformed learning.  Forthcoming in Brain and
Behavioral Sciences.

Darwin, C. (1871) The descent of man, John Murray Publications.

Dawkins, R. (1976) The selfish gene, Oxford University Press.

Dawkins, R. (1982) The extended phenotype, W. H. Freeman and Company.

Dennett, D. C. (1995)  Darwin's dangerous idea, Little, Brown 
and Company.

Donald, M. (1991) Origins of the modern mind, Harvard University
Press.

Dunbar, K. & Schunn, C.D.  (1990) The temporal nature of scientific
discovery: The roles of priming and analogy.  In: Proceedings of the
Twelfth Annual Meeting of the Cognitive Science Society, 93-100.

Durham, W.H.  (1991) Coevolution: genes, culture, and human diversity.
Stanford University Press.

Fehrer, E. & Raab D.  (1962) Reaction time to stimuli masked by
metacontrast.  Journal of Experimental Psychology, 63, 143-147.

Gabora, L. M. & Colgan, P. W.  (1990) A model of the mechanisms
underlying exploratory behaviour.  In The simulation of adaptive
behavior, eds. S. Wilson and J. A. Mayer, MIT Press.

Gabora, L. M.  (1995) Meme and variations: A computational model of
cultural evolution.  In 1993 Lectures in Complex Systems,
Addison-Wessley.

Gentner, D. (1983) Structure-mapping: A theoretical framework for
analogy.  Cognitive Science, 7(2).

Gick, M. L. & Holyoak, K. J. (1983) Schema induction and analogical
transfer.  Cognitive Psychology, 15, 1-38.

Gould, S. J. (1991) Bully for brontosaurus: reflections in natural
history.  W. W. Norton & Company.  p. 63-66.

Gould, S. J. & Vrba, E. S. (1982) Exaptation--A missing term in the
science of form.  Paleobiology, 8(1), 4-15.

Greene, P. J. (1978) From genes to memes?  Contemporary Sociology 7,
06-709.

Guilford, J. P. (1979) Some incubated thoughts on incubation.  Journal
of Creative Behavior, 13, 1-8.

Hebb, D. (1949) The organization of behavior, Wiley and Sons.

Hofbauer, J. & Sigmund, K. (1988) The theory of evolution and
dynamical systems, Cambridge University Press.

Hofstadter, D. R. (1985) Variations on a theme as the crux of
creativity.  In Metamagical Themas, 232-259, Basic Books.

Holgate, P. (1966) A mathematical study of the founder principle of
evolutionary genetic.  J. Appl. Probl., 3, 115-128.

Holland, J. K. (1975) Adaptation in natural and artificial systems,
University of Michigan Press.

Holland, J. H., Holyoak, K. J., Nisbett, R. E. & Thagard, P. R.  
(1986) Induction,  MIT 	Press.

Hopfield, J. J. (1982) Neural networks and physical systems with
emergent collective computational abilities.  Proceedings of the
National Academy of Sciences (Biophysics).  79(8), 2554-2558.

Hubel, D. H. & Wiesel, T. N. (1979) Brain mechanisms and
vision. Scientific American, 241(3), 150-62.

Hughes, D.M. & Nei, M. (1988) Patterns of nucleotide substitution at
major histocompatibility complex class I loci reveals overdominant
selection, Nature, 335, 167-170.

Hull, C. L. (1943) Principles of behavior, Appleton-Century- Crofts.

Kanerva, P. (1988) Sparse distributed memory, MIT Press.

Kaplan, N. L., Hudson, R. R., & Langley, C. H. (1989) The hitchhiking
effect revisited.  Genetics, 123, 887-899.

Karmiloff-Smith, A. (1986) From meta-processes to conscious access:
Evidence from childrenUs metalinguistic and repair data.  Cognition
23, 95-147.

Karmiloff-Smith, A. (1990) Constraints on representational change:
Evidence from childrenUs drawing.  Cognition 34, 57-83.

Karmiloff-Smith, A. (1992) Beyond modularity: A developmental
perspective on cognitive science, MIT Press.

Kauffman, S. (1993) Origins of order, Oxford University Press.

Keeler, J. D. (1988) Comparison between KanervaUs SDM and Hopfield-type
neural networks.  Cognitive Science 12, 299-329.

Klimesch, W. (1995) Memory processes described as brain oscillations.
Psycoloquy, 6.06.

Koestler, A. (1964) The act of creation, Picador.
  
Kojima, K. & Schaeffer, H. E. (1967) Survival processes of linked
mutant genes.  Evolution, 21, 518-531.

Kubose, S. K. & Umenoto, T. (1980) Creativity and the Zen koan.
Psychologia, 23(1), 1-9.

Langton, C. G.  (1992) Life at the edge of chaos.  In Artificial life
II.  eds. C. G. Langton, J. D. Farmer & S. Rasmussen, Addison-Wesley.

Lenat, D.B. (1974) The ubiquity of discovery.  Artificial Intelligence
Journal, 9, 257-86.
 
Luchins, A.S. (1942) Mechanization in problem solving.  Psychological
Monographs, 54, No. 248.

Lumsden, C. & Wilson, E. O. (1981) Genes, mind, and culture, Harvard
University Press.

Lynch, A. (1996) Thought Contagion: How Belief Spreads Through
Society, Basic Books.

Mandelbrot, B. B. (1982) The fractal nature of geometry.  W. H.
Freedman and Company.

Marr, D. (1982)  Vision, Freeman.

Maynard Smith, J. & Haigh, J. (1974) The hitchhiking effect of a
favourable gene.  Genetic Research, 23, 23-35.

McCulloch, W. S. & Pitts, W. (1943) A logical calculus of the ideas
immanent in nervous activity.  Bulletin of Mathematical Biophysics, 5,
115-133.

McFarland, D. J. & Sibly, R. M.  (1975) The behavioural final common
path.  Philosophical transactions of the London Royal Society 270B,
265-93.

Miller, G. A. (1956) The magic number seven, plus or minus two: Some
limits on our capacity for processing information.  Psychological
Review, 63, 81-97.

Minsky, M. (1985) The Society of Mind.  Simon and Schuster.

Nersessian, N. (1993) In the theoretician's laboratory: Thought
experimenting as mental modeling.  In PSA 1992, vol. 2, eds. D. Hull,
M. Forbes, & K Okrulik.  PSA.

Palmer, M. W.  (1992) The coexistence of species in fractal landscapes.
American Naturalist, 139, 375-397.

Pinker, S. (1995) The language instinct.  HarperPerrenial.

Plotkin, H. C. (1988)  The role of behavior in evolution.  MIT 
Press.

Popper, K. R. (1963) Conjectures and refutations, Routledge & Kagen.

Pribram, K. H. (1974) How is it that sensing so much we can do so
little?  In The neurosciences third sudy program, eds. F.O. Schmidt &
F.G. Worden, MIT Press.

Quillian, M. R. (1968) Semantic memory.  In Semantic information
processing, ed. M. Minsky, MIT Press.

Ram, A. (1993) Creative conceptual change.  Proceedings of the
Fifteenth Annual Conference of the Cognitive Science Society, 17-26.

Ray, T. (1991) An approach to the synthesis of life.  In Artificial
life II. eds. C.G.  Langton, C. Taylor, J.D. Farmer & S.  Rasmussen,
Addison-Wesley.

Robert, M. (1990) Observational learning in fish, birds, and mammals:
A classified bibliography spanning over 100 years of research.  Psych
Record 40, 289-311.

Rumelhart, D. E. & McClelland, J. L. eds. (1986) Parallel distributed
processing, Bradford/MIT Press.

Schank, R.C. (1983) Dynamic memory, Cambridge University Press.

Schank, R.C. & Leake, D.B.  (1989)  Creativity and learning in a 
case-based explainer.  	Artificial Intelligence, 40(1-3): 353-
385.

Scheuring, I. & Riedi, R. H.  (1994) Application of multifractals to
the analysis of vegetation pattern.  Journal of Vegetation Science, 5,
489-496.

Schuster, P. & Sigmund, K. (1983) Replicator dynamics.  Journal of
Theoretical Biology 100, 533-38.

Shannon, C.E. & Weaver, W. (1963) The mathematical theory of
communication.  University of Illinois Press.
  
Smith, W. J. (1977) The behavior of communicating, Harvard University
Press.

Spector, L. & Luke, S. (1996) Culture enhances the evolvability of
cognition.  In Proceedings of the 1996 Cognitive Science Society
Meeting.

Squire, L. R. (1992) Memory and the hippocampus: A synthesis from
findings with rats, monkeys, and humans.  Psychological Review, 99,
195-231.

Thagard, P. (1980) Against evolutionary epistomology.  PSA 1980,
ed. P. D. Asquith & R. N. Giere 187-96.

Tomasello, M. Kruger, A. C., & Ratner, H. H.  (1993) Cultural
learning.  Behavioral and Brain Sciences, 16, 495-552.

Van Loocke, P. (1991) Study of a neural network with a meta- layer.
Connection Science, Vol. 3, No. 54.

Von Neumann, J. (1966) Theory of self-reproducing automata, University
of Illinois Press.

Vygotsky, L. S. (1978) Mind in society: The development of higher
psychological processes, eds. M. Cole, V. John-Steiner, S. Scribner &
E. Souberman, Harvard University Press.

Wallas, G. (1926) The art of thought, Harcourt, Brace & World.

Walsh, R. & Vaughan, F. (1993) Paths beyond ego, Jeremoy P.
Tarcher/Perigee.

Weisberg, R.W. (1986) Creativity: Genius and other myths, Freeman.

Wright, S. (1969) Evolution and the genetics of populations,
University of Chicago Press.


FOOTNOTES

1. Note that by "causal principle" I mean something that generates
useful descriptions, rather than a law.

2. Though viruses are unique in the biological world in that they rely
on hosts to replicate, we will consider viral evolution an anomalous
offshoot of biological evolution, because: (1) the evolving patterns
of information are encoded as sequences of nucleotides, (2) variation
is through mutation and recombination, and (3) transmission and
selection are mediated through genotype.

3. Or one could argue that the 'selection' of matter over antimatter,
and its subsequent amplification and variation, constitutes yet
another form of evolution.

4. This example is a variation of one transmitted by R. Boyd
[pers. com.]