Imaginative Immersion and Structural Learning
24 July 2000

Imaginative Immersion

Semantic and episodic knowledge stored in long-term memory (LTM) is organized into extensive propositional networks.
Exposure to one cue that retrieves information leads to the recall of related knowledge through the process of spreading
activation.

3 strategies function to store and organize information in LTM:

1.Maintenance Rehearsal – The strategy most researchers are familiar with and the definition of rehearsal. This strategy
involves the mere repetition of information (e.g., phone numbers, definitions for multiple choice tests, names, etc.). Unless
this information is used extensively, retrieval is unlikely to occur as this strategy constructs "islands" in the organization of
LTM. Information learned via maintenance rehearsal must be practiced to the point of becoming automatic or
procedural. In short-term/working memory, this strategy functions by re-entering information, which increases the
stimulation of the neurons in the reverberating circuit. If interference is not encountered during this process and
information is practiced extensively to the point of automaticity, retrieval is possible. Mature maintenance rehearsal
strategies begin to be used by children around 6 years of age.

2.Elaboration – A more advanced strategy for storing and organizing information in LTM. The procedure is to connect
new information to prior knowledge, which leads to the creation of propositional networks complete with "bridges
connecting islands." New information is deeply processed and made meaningful through elaboration. Neurologically, this
new information becomes a part of a well-worn neuronal pathway. These connections provide the roads for spreading
activation to occur. This strategy is not used spontaneously by children until close to adolescence.

3.Imaginative Immersion – We hypothesize the presence of a third strategy for storing and organizing information in LTM.
This strategy is used extensively in the learning mode and requires rich mental simulations. Agent register swapping is also
necessary for imaginative immersion in fictive learning. This strategy is employed in learning through pretense, fiction or
pretend play, and when mentally rehearsing to perform a skill (e.g., yoga asana, basketball plays, rock climbing, skiing,
etc.). In using this strategy, the behaviors practiced are embodied and involve some form of enactment, whether explicitly
with the body as in pretend play or mentally as in fictive learning and mental rehearsal of a skill. Whereas elaboration is a
completely mental activity of connecting new and prior knowledge. Imaginative immersion, as stated above, can take
place during three activities:

a.Pretend Play – Explicit enactment due to working memory constraints. Children need props (even if symbolically
mapped onto abstract environmental affordances) and other children to distribute the cognitive load. Embodied in
the fact that while rehearsing cultural skills and knowledge or even predator evasion they ARE what they pretend
to be. Even in the rehearsal of mindreading strategies ("Pretend I ____ and you didn’t know.") the child assumes
mental states that are counterfactual but can be imagined and played out with entailments ("Mommy" gets mad
because she should have been told "Baby" was going out to play.) In this way, children can explore the possible
outcomes of behaviors as interpreted by another with a specified epistemic state. The playmate gains knowledge
about possible reactions to behaviors by virtue of possessing a certain epistemic state.

b.Fictional Learning – Children’s books are illustrated to distribute cognition. They provide cues relating to the story
characters and the plot. As WM capacity expands, the story becomes increasingly simulated mentally without
memory aids. Imaginative immersion is still taking place but the enactment of the story is no longer physically
necessary – fully internalized immersion. The swapping of agent memories and the full immersion into a mental
simulation supplemented by one’s own idiosyncratic memories are indicators of the embodied nature of rehearsal
of information/strategies through this form of learning occurring in the learning mode. Imaginative immersion may
also be in action when listening to other people tell stores about their lives. Just imagine sitting around a campfire
after a hunting party returns from a hunt. Warm, safe, fed, you listen and watch the symbolic enactment of the hunt
that has been adapted to the particulars of the most recent success. Imaginative immersion in fictional learning in
action. Unconscious changes in probability weightings and confidence levels for strategies to coordinate a hunt, to
successfully collaborate with others, predation strategies, etc.

c.Mental Rehearsal of Skills – Talented athletes run mental simulations of skills when not able to engage in actual
physical practice. This behavior has been documented to be related to improvements in ability as the skills are
performed with perfection in the imagination.

When I was a caterpillar, I ate bugs

This kind of agent pretense is incongruous in light of the other two theories we have of the phenomenon; on the other hand, it
provides a potential link between them. Freedom is clearly swapping agent registers, yet he does not benefit from this in either
of the two ways outlined above: this is neither motor practice nor mind-reading practice. It looks like he's using agent pretense
to access world knowledge and store it as relevant information.

We developed the very tentative hypothesis that this constitutes a genuine third function of agent pretense: one that can be
undertaken consciously in individual play. The function is to gather information about the environment and store it effectively, so
that it can be accessed. The suggestion is that it is difficult to create appropriate mental spaces for storing information about the
environment. One way to create effective mental spaces -- to organize the knowledge dynamically -- is to imaginatively project
oneself into a different perspective, such as that of a caterpillar. This creates a cognitive frame that is sufficiently motivated to
gather and store information.

This view of agent pretense as a strategy for obtaining and storing information about the environment has interesting
implicaitons. It is based on the intuition that in the executive mode, animals have robust adaptations for picking out only that
information in their surroundings that is relevant to them. They don't have the cognitive resources to consider other perspectives,
other Umwelts. Memory systems have been carefully engineered to screen out information that is not relevant to self; there is a
strong bias against irrelevant information.

Human beings, however, at some point entered what has been called the cognitive niche. This means that the kind of
information that was previously irrelevant became in effect a possible resource. We evolved a terrific learning capability. The
basic motivational structure of the brain, however, is unchanged. There is still a bias against information that is not clearly
relevant. In the cognitive niche, this constitutes an adaptive problem: we need to find ways of exploiting what instinctively may
appear to be useless and irrelevant.

There may be a devious way of circumventing this motivational limitation -- a strategy, as it were, that the implicit curriculum
can activate to skirt the epistemic censors of the self. It is to project oneself into the agent memories of another, to adopt
another perspective. The older adaptations kick into gear: if you think you are a bear, a lot of things suddenly become relevant
and salient. The highly sophisticated information gathering devices of the mind, including their motivational component, can as it
were be coopted to gather all kinds of information that the literal or realistic self would disregard. It might become important to
imagine and remember what caterpillars might eat -- even though they don't, in fact, eat bugs. Mental spaces can be created
that are effective in organizing information in an integrated manner: for instance, information about what caterpillars eat might be
accompanied by the actual motions of moving forward, catching the bug, and eating it. In this way, memories are encoded in a
context that makes it easier to retrieve.

This model is broadly mappable onto cognitive theories of learning that speak of spreading activation and the formation of
propositional networks. However, this particular view -- which could easily be misguided -- adds an evolved and
domain-specific dimension. It is because we are embodied that we are constrained to learn through imaginative immersion; our
ancient cognitive systems for information filtering and retention have evolved to define relevance with respect to self. The
implicit curriculum, through the use of pretense, tricks the mind into transcending its own pettiness. In the context of literature,
imaginative immersion through fiction provides the only effective way to learn about life in social situations you cannot directly
experience. By imagining you are a king, or the daughter of the Earl of Berkeley, you gain access to, become motivated to
explore, and retain in a tidy way the information about that society as if it were relevant to the self. Fiction in this view is a
powerful technology for tapping into the resources of the mind and breaking open its primitive and constricted mold.

It is worth noting that this kind of fiction also involves a form of mindreading, and that this function of fiction is unlocked by
mindreading. Its proper domain may be the ability to adopt another's perspective when the other is telling a story: you
imaginatively project yourself into the deictic center of the story. The argument presented here simply seeks to suggest why this
is such an effective learning strategy. It is not simply to train mind reading, but to gain access to world knowledge. But it is
made possible by mindreading and undoubtedly trains it. It provides, in a sense, an incentive to read minds: this is why
mindreading matters. By reading minds -- and specifically, by adopting their perspectives -- you can gain access to information
that otherwise would have been classified as irrelevant. It's somewhat unclear if this kind of learning through fiction is really
different from the third function, to train mindreading. It may be possible to combine them and simply say, by training
mindreading through agent pretense, the child (or adult) becomes motivated to learn about the world from a different
perspective -- information that may be useful later on. This perspective supplements the mindreading perspective: mindreading
opens up a vast possibility space not only because information becomes available from other minds (strictly speaking, classic
mindreading is not required for this), but because the mind becomes motivated to acquire, organize, and store information as if
it were relevant.

What would happen if I asked Freedom, "Were you really a caterpillar?" Would he say yes? Do we need to deny our state of
pretense even though we know this to be the case? Then ask Freedom the same question when he is not in the midst of
pretense, what would the answer be? Possibly, keeping track of the fact that something is pretense may fill needed working
memory space. Mary Hegarty says people do not even have enough working memory capacity to handle large novels like War
and Peace. We possibly free working memory capacity by placing ourselves in situations where we do not have to worry
about – track – reality. With history/fiction, we must deny pretense is taking place to get a really good experience of immersion
– a detailed simulation. If kids say they are not pretending, the same phenomenon may be taking place. Maybe the 1st person,
self pretend, format more reliably stores newly learned information.

Structural Learning

Chase Play Project

Conscious Structural Learning
Stephanie and Francis
11 June 2000

I. Classic Chase

EA values are the default VA values; this is the simplest case. Transformations are based on affordances and mappings are
based on the similarities between the perceived and imagined objects. Re-routing is a simple metaphorical mapping. (Srini
Narayan -- pretense and metaphorical mapping)

II. Narrative Chase

EA values still default for VA values. Transformations occur, in part, via affordances through mapping. Memories are recruited
to build VA by specifying environmental and role values. Still using re-routing but play is enriched through the recruitment of
cultural schemas.

III. Cultural Play

Developmental trajectory of pretend play in consideration of changes in working memory, the ability to map, and build and use
enriched cultural schemas:

1.can map only one object in an object substitution = little WM capacity; basic, barebones schemas
2.can ascribe active agency (e.g., doll can be hungry) = more WM and more knowledge base
3.can ascribe passive agency (e.g., doll can sleep)
4.role play = many mappings to environment and self; need more WM; larger cultural schemas
5.narrative play = many mappings, tracking coherent story; enriched schemas; mindreading
6.sociodramatic play = many mappings, tracking story; need large knowledge base; mindreading; tracking other's
representations of scenario (executive mindreading)

III. The Novel

Still need to suspend EA content; this may be more crucial as content may be entirely irrelevant to VA. The words read
function as retrieval cues to recall personal memories. This retrieval process may require inferences to be drawn as many details
are left tacit in the story (to specify which of many interpretations should be retrieved, i.e., lexical access). These memories
constitute environmental values which in pretend play are imported from the EA.

What is the relation between the VA and the EA for structural learning?

Because the VA is nested in EA, a possibility for continual analogical mappings is created. Continually mapping onto your own
agency may have a more immediate effect on self-understanding. This will only occur if relavance criteria are met, meaning there
is a lot of overlap between traits, context, and possibility spaces of VA and EA (e.g., Tom Robbins novels; the Gladiator film).
When pretense arouses emotions, you're actually experiencing those emotions even when the VA disintegrates. Now you must
explain the presence of the emotions. The emotions are still liked to the areas of your possibility space just explored by the
novel. Behavior can be immediately influenced. The emotion is in full figure as are the strategies leading to that feeling. If there is
a positive affect, those strategies will be used in the present; if there is a negative affect, one will be reluctant or resistant to
employ those strategies.

This process of conscious structural learning can also happen in children's pretend play. For example, Michael was playing
committee meeting with his friends. He may now feel more confident after playing the leader role and will want to exercise
leadership in the executive mode. He may also extract general properties for use in social relations.

Structural Learning:

1.Unconscious Structural Learning -- details vanish (all that was in consciousness); may have vague awareness that is
nonverbal but information learned was passed just under conscious awareness to affect probability weightings and
confidence levels of strategies (e.g., can be resistant to a message in lecture but learn it through story -- Parables)
2.Conscious Structural Learning -- The connector between the VA and EA is emotion due to meeting relevance criteria
(see diagram). The emotion lingers and because you've been mapping those aspecs of the EA are in figure: the
characteristics are maintained through analogical mapping during immersion; the possibility space explored remains in full
figure after the virtual world has collapsed due to the emotional level. The VA is still resonating because it is still tied to
the possibility space. This will dramatically alter probablility levels and confidence levels of strategies.

Example presented in fiction: Fannie Flagg Fried Green Tomatoes -- in frame, woman visiting the elderly woman to hear the
stories of the two heroines from the past begins to assert her agency in her present. TOWANDA!!!!

Structural learning: In implicit pedagogy, a process of exploration (1), leading to the development of strategies (2), which are
assigned levels of confidence and probability of success (3).

1.The development of a new skill creates a new possibility space for action, defined both by the skill itself and by the
environment in which it is developed
2.This possibility space is exploited to develop coherent strategies of action -- strategies that are tailored to the kinds of
challenges the environment presents
3.Practice causes largely unconscious changes in the distributions of associations that regulate the choice of strategies
(Siegler 1986)

New skills may be specified as target values of evolved adaptations; they can also be the result of schematization.

Schematization: in learning, a largely unconscious process of reductive abstraction that maintains appropriate structural
invariances within experiential content; it results in image schemas (cf. Mandler 1992) and scripts. Schematization frees up
working memory, which allows structural learning to take place.

Script: in Schank (1990), a set of tacit instructions for appropriate actions and expectations in a particular setting, such as a
restaurant; in implicit pedagogy, scripts are the result of schematization.

Structural learning from fiction

This extended model of structural learning has some interesting features. First, there is the idea that new skills may be specified
as target values of evolved adaptations or they can be the result of schematization. Schematization is proposed to be a mode of
learning that chunks episodic memories into compact procedural or declarative memories, forming image schemas and scripts.
The formation of a new schema frees up working memory, which allows an awareness of the possibility space opened up by
the new skill. This in turn allows the exploration and discovery of new strategies -- structural learning.

Fictive as well as historical narratives can be helpful in exploring the possibility space of a new skill by developing a variety of
socially embedded strategies. The structural learning hypothesis of fiction states that even when stories are fictional, they can
lead to largely unconscious changes in the distributions of associations that regulate the choice of strategies in real life

Structural Learning: A Sociological Approach

McClelland (1961, 1965) collected myths, folklore, and children's stories from many countries and applied methods of content
analysis to determine levels of achievement motivation portrayed in stories. In a study of preliterate societies, he analyzed oral
folktales for achievement imagery. Twenty-two cultures were determined to have a high need for achievement imagery in their
stories. 75% of these cultures had individuals who could be classified as entrepreneurs. Twenty-three cultures low in need for
achievement imagery and only 33% had individuals who could be classified as entrepreneurs. Cultural attitudes about
achievement as measured in stories are correlated with number of individuals willing to take risks of managing a business that
has the potential of increasing the economic health of the culture. In 1965 he demonstrated that men high in need for
achievement tend to select entrepreneurial occupations.

He also analyzed children's beginning readers from several countries used in 1925 on need for achievement. He used electric
power consumption figures from 1925 to 1950 as a measure of economic growth to correlate with the analysis of stories. For
78% of countries above mean for achievement, there were large increases in energy consumption. For those below the mean,
only 25% showed similar increases. There was no correlation between energy consumption from 1925-1950 and children's
readers from 1950. He concluded from this finding that children's literature can be used to predict future economic changes for
a nation.

Bradbern and Berlew (1961) analyzed English lit from 50 year intervals between the years of 1400 and 1830. They use coal
imports as their measure of economic growth. Achievement images in stories appear about 50 years before economic changes.
When imagery was high, 50 years later coal imports were higher and vice versa. They postulate that achievement images in
popular literature precede economic changes possibly due to influencing attitudes about achievement which in turn influences
entrepreneurial behavior of large groups of people when they become adults.

De Charms and Moeller (1962) analyzed children's readers from the US from 1800 to 1950 and illustrated that achievment
imagery peaked around 1890. Peak economic growth would be predicted to occur around 1940 (following the 50 year trend).
The decline in achievement imagery from 1890 to 1950 in stories would predict, if the correlation holds, that the US is now on
a downward trend.

Word of caution: Remember correlations do not indicate causation and the relationship is not a perfect correlation of +1.0.
Some countries low on achievement imagery experienced economic growth and vice versa.

Data on what children are learning

Basic model = target value opens possibility space which is used to explore strategies that are motivated by the executive
mode, real world desires. Learning consists of unconscious changes in the confidence levels associated with strategies
and the probability levels of each solution in relation to each strategy resulting in peaked or flat distributions of
association that influence strategy choice (Siegler, 1986) when confronted with the real situation or during practice in
pretense.