1. EP1.
Learning
Elena Pasquinelli
Educa3on, cogni3on, cerveau
Cogmaster 2010‐2011
2. Op3miza3on of educa3on
• “Considera3ons on the op3miza3on of
educa3onal strategies should take into
account knowledge on brain development and
learning mechanisms that has been
accumulated by neurobiological research over
the past decades.” (Singer, in BaKro, Fischer &
Léna, 2008, p. 97)
3. • Pre‐programming and experience: rela2onship
between learning, development, evolu2on
• Learning all life‐long?
• Learning everything?
• Prac2cal issues related to plas2city and learning
5. Defini3on of learning
• Learning = modifica3on of stored • “any learning, i.e. the
knowledge and of computa3onal
programs
modifica2on of
computa2onal
• Which takes place through the programs and of stored
modifica3on of the brain knowledge, must occur
func3onal architecture through las$ng
changes in their
• Learning = long‐las3ng change in func$onal
the func3onal architecture of the
brain
architecture.” (Singer,
2008, p. 98)
6. Defini3on of knowledge
• Knowledge is the product • « there is no dichotomy between
of biological processes, hard‐ and soXware in the brain. The
which determine or way in which brains operate is fully
modify the func3onal determined by the integra3ve
architecture of the brain proper3es of the individual nerve
• Learning is one of these cells and the way in which they are
processes interconnected. It is the func3onal
architecture, the blueprint of
connec3ons and their respec3ve
weight, that determines how brains
perceive, decide, and act.
• … all the knowledge that a brain
possesses reside in its func3onal
architecture. » (Singer, 2008, p. 98)
7. Modifica3on of the brain’s func3onal
architecture: 3 processes
• 3 different “altering the integra2ve proper2es of
processes are individual neurons…
responsible of changing the anatomical connec2vity
the paRerns, …
specifica3on/ modifying the efficacy of excitatory
modifica3on of and/or inhibitory connec2ons.
the brain’s
…”(Singer, 2008, p. 98)
func3onal
architecture
(and thus, of
knowledge “Evolu3on,
acquisi3on): Ontogene3c development,
And learning.” (Singer, 2008, p. 98)
8. a. Learning and evolu3on
• Evolu3on has selected both learning mechanisms
and knowledge contents:
– E.g. : “Fire together, wire together”
– E. g. : How to evaluate regulari3es, extract rules,
associate signals, iden3fy causal rela3ons,
reason, associate emo3ons to sensory s3muli
– E. g. How to interpret sensory signals
• The brain stores knowledge even before making
experiences: it’s not a tabula rasa.
– Educa2on cannot be considered as the task of
filling a hollow box
9. b. Learning and development
• The brain at birth is s3ll immature: neurons are in place, basic
distant connec3ons between neurons are formed, but not the most
part of the neurons of the cortex
• From birth to the end of puberty, neural circuits are formed and
selected
– Development includes 3me window, or expects certain s3muli at specific
periods of the life of the animal in order to implement certain func3ons
• During development connec3ons are formed and tested (“fire
together‐wire together”): those connec3ons, which have a high
probability of being ac3vated simultaneously are consolidated,
those which have a low probability are discarded.
• AXer birth, this networking ac3vity is influenced by individual
experience of the environment and sensory signals
10. c. Learning (adult) = func3onal modifica3ons of
brain’s func3onal architecture
• Development and learning cross their paths, but aXer puberty neural
circuits and the structural architecture of the brain are (apparently)
mostly stabilized
• Adult learning: Func3onal modifica3ons
– strength of the connec3ons,
– efficacy of the connec3ons
• are the main mechanisms for the modifica3on of the func3onal
architecture of the brain
• Learning does not modify the architecture of the brain at a structural
level (mostly):
• it produces func3onal modifica3ons that affect the strength of the
connec3ons between neurons (synapses) = Func2onal plas2city
11. The role of experience
• In addic3on to gene3c mechanisms, the brain is
modified by experience
• Both
– At the level of epigenesis and development (see
effects of depriva3on)
– At the level of learning
• But with
– constraints to what can be learnt:
• certain mechanisms protect the brain from
adap3ng to any new informa3on coming from
the environment
13. Cri3cal (sensi3ve) periods for learning
• Cri3cal periods = 3me‐
window opportuni3es
• Development of vision
– Hubel & Wiesel, 1970:
monocular depriva3on
reduces the number of cells
responding to the ac3vity of
the deprived eye
– monocular depriva3on has
different effects at different
ages
• Development of language
14. The myth of the first three years
• The no3on of cri3cal periods has been
domina3ng the world of educa3on and has
given birth to myth of the first three years
• Bruer, 1997 describes this myth as a typical
case of bad transla3on from
neuroscien3fic data to educa3onal
applica3ons
• Bruer, 1997 cri3cizes the iden3fica3on of
learning with synaptogenesis:
– Different systems have different sensi3ve
periods, in the sense that they do not develop
at the same rate (including within the visual
system)
– Human cri3cal periods are not necessarily the
same as animals
– The brain is more plas2c than accorded
before
– Learning cannot be reduced to
synaptogenesis
15. general rule for neuroeduca3on
• Bruer has used the myth of the first
three years for showing that
neuroscience is s3ll a bridge too far
from educa3on, and can give rise to
neuromyths and misapplica3ons. So
pay aKen3on to:
• generaliza3on of considera3ons that
are extracted from
– Animal experiments
– Data on specific func3ons
• erroneous iden3fica3on of brain
mechanism and behavioral
phenomenon
– E.g., Iden3fica3on of learning with
synaptogenesis
16. From cri3cal periods to different forms
of plas3city
• (Greenough, Black & Wallace, 1987) have introduced the dis3nc3on between
two ways in which experience modifies the brain:
• Experience‐expectant plas2city:
– Selected by evolu3on
– Concerns sensory motor func3ons
– Allows to fine‐tune the sensory motor systems in rela3onship to the environment
– Through the selec2on of synapses that have been generated in excess
– Defines the s3muli that should be found in the environment for the func3on to develop in a
certain way
– Experiences are very general and concern s3muli, which are normally present in the
environment
• Experience‐dependent plas2city:
– Does not depend on mechanisms that have been selected by evolu3on according to a precise
3ming
– Evolu3on has selected a capacity to learn from experience in general
– Through the genera2on of synapses, and the modifica2on of the strength of the synapses
17. 3 mechanisms for func3onal and
structural plas3city
• Plas3city is the basis of learning from • « The most fascina3ng and important
experience property of mammalian brain is its
• 3 mechanisms: remarkable plas3city, which can be
– Synap3c plas3city = change in strength thought of as the ability of
or efficacy of synap3c transmission experience to modify neural circuitry
– Synaptogenesis & synap3c pruning and thereby to modify future
– Excitability proper3es of single neurons thought, behavior,
feeling.» (Malenka, 2002, p. 147)
• Synap3c plas3city can be transient
(short term phenomena such as
short‐term adapta3on to sensory
inputs) – depends on modula3on of
transmiKer release
• Or long las3ng: long‐term form of
memory
– LTP/LTD (long‐term poten3a3on/long‐
term depression) mechanisms
18. LTP
• LTP: repe33ve ac3va3on of excitatory synapses in the hyppocampus
causes an increase in synap3c strength that can last for hours
• LTP is hypothesized to be involved in the forma3on of memories and
more generally in informa3on storing, hence in learning in general,
because LTP and learning considered at the behavioral level share
some proper3es:
– LTP can be generated rapidly and is prolonged and strengthened by
repe33on
– It is input specific (it is elicited at the ac3vated synapses and not at adjacent
synapses of the same neuron)
– It’s long‐las3ng
• How? Modifica3on of dendri3c spines? Growth of spines? Genera3on of new
synapses as a consequence of the splinng or duplica3on of exis3ng spines?
• Incorpora3ng structural changes into the mechanisms of long‐term synap3c plas3city
provides means by which the ac3vity generated by experience can cause long‐las3ng
modifica3ons of neural circuitry
19. More “structural” plas3city
• “ Un3l rela3vely recently, it was widely assumed that, except for certain cases of response to brain
damage, the brain acquired all of the synapses it was going to have during development, and that
further plas3c change was probably accomplished through modifica3ons of the strength of
preexis3ng connec3ons.
• … it has now become quite clear that new connec3ons may arise as a result of differen3al housing
condi3ons and other manipula3ons throughout much, if not all, the life of the rat…
• There has not yet been a specific demonstra3on of what might be represented by the changes in
synap3c connec3ons brought about by differen3al environmental complexity, nor are the details
of the rela3onship between brain structure and behavioral performance.” (Greenough, Black &
Wallace, 1987, p. 547‐548)
• “However, there are a few excep3ons. Over the past years, evidence has become available that in
a few dis3nct brain region, parts of the hippocampus and the olfactory bulb neurons con3nue to
be generated throughout life, and these neurons form new connec3ons and become integrated in
exis3ng circuitry.”
• “Thus in these dis3nct areas of the brain, developmental processes persist throughout
life…” (Singer, 2008, p. 108)
20. Structural plas3city in the adult brain
• MRI of licensed London taxi drivers were analyzed and
compared with those of control subjects who did not drive
taxis.
• The posterior hippocampi of taxi drivers were significantly
larger rela3ve to those of control subjects.
• Hippocampal volume correlated with the amount of 3me
spent as a taxi driver (posi3vely in the posterior and
nega3vely in the anterior hippocampus).
• These data are in accordance with the idea that the posterior
hippocampus stores a spa3al representa3on of the
environment and can expand regionally to accommodate
elabora3on of this representa3on in people with a high
dependence on naviga3onal skills.
• It seems that there is a capacity for local plas3c change in
the structure of the healthy adult human brain in response
to environmental demands. (Maguire, et al.,2000)
22. The role of educa3on
• 3 possible views:
– One can learn everything, and learns it from scratch
– What we learn depends on past experiences and is
constructed star3ng from these experiences, but one can
learn everything
– The way brain has been shaped by selec3on strongly
constrains what can be learnt
• (Posner & Rothbart, 2007)
23. Can we learn anything? Constraints and biases
• Learning experiences sculpt the brain
and cons3tute a framework for future
learning
• E. g. According to Kuhl (2004) mother
language learning builds a mental filter
that limits second language learning
• the “cri3cal period” depends on
experience as much as 3me, and is a
process rather than a strictly 3med
window of opportunity that is opened
and closed by matura3on.
– (Bransford, et al, in Sawyer, 2009)
24. Can we learn anything? Evolu3on and selec3on
• «… I have oXen observed that educators hold an implicit model of brain as a
tabula rasa or blank slate (Pinker, 2002), ready to be filled through educa3on and
classroom prac3ce. In this view, the capacity of the human brain to be educated,
unique in the human kingdom, relies upon an extended range of cor3cal plas3city
unique to humans. The human brain would be special in its capacity to
accommodate an almost infinite range of new func3ons through learning.
• In this view, then, knowledge of the brain is of no help in designing educa3onal
policies.
• …. Much of current classroom content, so the reasoning goes, consists in recent
cultural inven3ons, such as the symbols we use in wri3ng or mathema3cs. Those
cultural tools are far too recent to have exerted any evolu3onary pressure on brain
evolu3on. … Thus, it is logically impossible that there exist dedicated brain
mechanisms evolved for reading or symbolic arithme3c. They have to be learned,
just like myriads of other facts and skills in geography, history, grammar,
philosophy … The fact that our children can learn those materials implies that the
brain is nothing but a powerful universal learning machine. » (Dehaene, in BaKro,
Fischer, & Léna, 2008, p. 233).
25. Biology and culture
• Implica3on of
the idea of
tabula rasa: each
learner is
radically
different from
other learners,
and the same
cerebral areas
can be affected
to different
func3ons
26. Neural recycling hypothesis
• “… Close examina3on of the func3ons of those brain areas in
evolu3on suggests a possible resolu3on of this paradox. It is not
the case that those areas acquire an en3rely dis3nct, culturally
arbitrary new func3on. Rather, they appear to possess, in other
primates, a prior func3on closely related to the one that they will
eventually have in humans. … rela3vely small changes may suffice
to adapt them to their new cultural domain.
• « neural recycling hypothesis », according to which the human
capacity for cultural learning relies on a process of pre‐emp3ng or
recycling pre‐exis3ng brain circuitry.
• In my opinion, this view implies that an understanding of the
child’s brain organiza3on is essen3al to educa3on.
27. Neural recycling & mathema3cs
• Arabic digits and verbal
numerals are culturally
arbitrary and specific to
humans
• the sense of numerical
quan3ty is not: it is
present in infants and
animals
• We learn to give meaning
to our symbols and
calcula3on by connec3ng
them to this pre‐exis3ng
quan3ty representa3on
28. Neural recycling & reading
• Visual cortex presents mechanisms for invariant shape recogni3on
• Visual cortex is connected with auditory and seman3c areas
• Visual cortex responds to T shapes, circles, superposed circles..
• Many of these shapes resemble to our leKers
• We do not need to create a reading area ex novo, but can preempt
other visual and auditory mechanisms
30. From theory to prac3ce
• How can we generate successful interven3ons for
promo3ng relevant learning ?
– How do we pass from theory to prac3ce?
– Which kind of theory and evidence do we need?
– What is relevant learning?
– Learning that is long‐las3ng and transferable
– How do we promote learning that is long‐las2ng and
transferable?
31. Plas3city in prac3ce
• “Learning and brain plas3city are fundamental
proper3es of the nervous system, and they hold
considerable promise when it comes to learning a
second language faster, maintaining our perceptual
and cogni3ve skills as we age, or recovering lost
func3ons aXer brain injury.
• Learning is cri3cally dependent on experience and the
environment that the learner has to face.
• … we are s2ll missing the recipe for successful brain
plas2city interven2on at the prac2cal
level.” (Bavelier, et al., in Gazzaniga, 2009, p. 153)
32. Training & Relevant learning
• In many cases, training produces
effects that cannot be considered
as relevant learning, because
(Bavelier, et al., in Gazzaniga, 2009,
p. 153):
– They are not long‐las2ng : an effect
on learning is not proved by
experiments that evaluate short‐
term effects (e.g.: violent effects of
violent video games)
– They are not sufficiently
generalized: an effect on learning
that is bound to the trained task is
barely interes3ng
– Other variables than the learning
experience produce an effect, but
are not controlled for and evaluated
34. Learning as reusable
• “Learning involves acquiring new informa3on • Learning is supposed to be re‐usable
and u2lizing it later when necessary. Thus, – An example: Imagine a motor therapy which
any kind of learning implies generaliza2on of induces the learning of new movements, but
the originally acquired informa3on: to new these movements can only be accomplished
occasions, new loca3ons, new objects, new in the therapy room
contexts, etc. However, any piece of new
informa3on that an organism perceives is
episodic and par3cular: it involves a single
3me, a specific loca3on and context, and
par3cular objects).” (Gergely & Csibra, 2009,
p. 3)
• “The ques3on of how one can learn (i.e.,
acquire general knowledge) from bits of
episodic informa3on is known as the
induc2on problem and has been tackled by
various theories of learning. These usually
rely on sta3s3cal procedures that involve
sampling mul3ple episodes of experience to
form the basis of generaliza3on to novel
instances.” (Gergely & Csibra, p. 3)
35. The neuromyth of the Mozart effect
• The Mozart effect
(Rauscher, Shaw, Ky, 1993)
– Effects of listening music on
spa3al reasoning (Stanford‐
Binet IQ test)
– Enhancement
– For 15 minutes
• a classic case of
performance enhancement
that is NOT a form of
learning, because it does
not last
• … and a classic neuromyth
– listening to Mozart increases
the IQ
37. Aggression and violent video games
• Violent video games
seem to produce
effects on
physiological
arousal, verbal
violence, but these
effects are only
tested few minutes
aXer the exposi3on
(Bavelier, et al., in
Gazzaniga, 2009, p.
154)
38. Transfer
• “In the field of learning, transfer of
learning from the trained task to even
other very similar task is generally the
excep3on rather than the rule.
• For instance, Pashler and Baylis (1991)
trained subjects to associate one of
three keys with visually presented
symbols (leX key = P or 2, middle key =
V or 8, right key = K or 7). Over the
course of mul3ple training blocks,
par3cipants reac3on 3me decreased
significantly. However, when new
symbols were added that needed to be
mapped to the same keys in addic3on
to the learned symbols … no evidence
of transfer was evident.” (Bavelier, et al.,
in Gazzaniga, 2009, p. 153‐154)
39. Methodological issues
• Studies on the effects of training on
learning should prove that the effects are
long‐las3ng and that there is a causal
rela3onship between the kind of training
and the learning effect (Bavelier, et al., in
Gazzaniga, 2009, p. 154‐155)
– The Hawthorne effect of learning:
mo3va3onal factors influence
performance, but they are not part of
the learning experience being
evaluated
– The popula3on effect: causal links are
not the same than correla3ons, since
correla3on could depend from
external factors