Implicit Learning processes in Children

Implicit Learning processes Implicit Learning processes in Children in Children

Annie VINTERAnnie VINTERL.E.A.D/CNRSL.E.A.D/CNRS

University of BourgogneUniversity of Bourgogne

Implicit learningImplicit learning

1. Definition of implicit learning 2. How to study this learning phenomenon in

laboratory settings? 3. A developmental approach 4. About the experiments currently in progress… 5. Conclusion: a useful learning mode?

Learning ?Learning ?

Intentionally, under the supervision of a « teacher »:

Without intention, independently of any « teacher »:

Explicit Learning (or conscious)

Implicit Learning (or unconscious)


1. Definition.

Implicit Learning ?Implicit Learning ?

Designates the type of learning which underlies all the capacities that people acquire spontaneously, without the help of any teacher or supervisor. For instance,

*mother tongue*basic knowledge on the world’s physical laws* Major part of motor behaviors* Part of social behaviors* Some aspects of written language* Some knowledge on music


When an individual’s behavior becomes sensitive to the structural features of a situation which occurs repeatedly, without this sensibility being due to an intentional use of explicit knowledge on the situation.

* 1. Occurs independently of any intention to learn, each time we are repeatedly confronted with situations that present stable regularities.

* 2. The direct interactions between the learner and the situation provoke a progressive and adapted transformation of his behavior, in a way that « resonates » with the relevant characteristics of the situation.

Implicit learning ?Implicit learning ?

• A comment:

What is crucial is to define what are the « relevant characteristics » or the « structural features » or the « regularities » of the situation: what does the individual really learn implicitly in such a condition ?


As a matter of fact, it is generally very difficult to determine what is learned implicitly, above all in the natural learning situations (daily life conditions).

There are always several possibilities:

- Abstraction (unconscious) of rules?- Memorisation of similar cases or events?


That’s one of the reason why the study of these implicit learning processes is simpler to carry out in laboratory situations (using artificial paradigms).


2. How to study this learning phenomenon

Laboratory (experimental) situationsLaboratory (experimental) situations

They constitute miniature situations of natural situations: the duration of learning is drastically shorter (usually, less than 1 hour); the complexity of the material to learn is much reduced.

In these situations, the structure of the material processed by the subjects is artificially built by the experimenter.

Furthermore, to mimick a learning condition without any intention to learn, the subjects are never informed on the structure of the situation they are confronted with.

Numerous experimental paradigmsNumerous experimental paradigms

Task of serial reaction times (SRT): Nissen & Bullemer, 1987….

Task of artificial grammar: Reber, 1967…. Task of Kushner et al. (1991)Task of Kushner et al. (1991) Task of incidental covariationsTask of incidental covariations: Lewicki,

1986... Control of dynamic informatical systems :

Berry & Broadbent, 1984. And othersAnd others……

The classical paradigmsThe classical paradigms

The Serial Reaction Times paradigm (SRT): Nissen & Bullemer, 1987.

…DCDABDBDACDBADCACABDACBDACDBADCABACD

A…

Exemple:

Serial Reaction TimesSerial Reaction TimesA

B

C

D

The mean reaction times associated to the repeated sequence become significantly faster than those associated to the random sequences, as if the subjects had learned the regularity of the sequence.

Simultaneoulsy, the subjects are unable to say anything about the repeated sequence.

SRT: what do we observe ?SRT: what do we observe ?

The classical paradigmsThe classical paradigms

The Serial Reaction Times paradigm (SRT): Nissen & Bullemer, 1987.

The Artificial Grammar Paradigm: Reber, 1976.

Grammar used in Peigneux, Meulemans, Van der Linden, & Petit, 1999).

IN

M

X

M

V

R

X

V

V

M

R

R

T

OUT

OUT

X

R

Artificial GrammarArtificial Grammar


* Study phaseMXRMVRVMVMRMXTRVXVRMXTMVXTR ……

* Test phase: Is the item grammatical or not ? VXV ?

MXTR ?VRMTX ?……

While the subjects have the feeling that they respond randomly, they reply more often « grammatical » to an unseen grammatical item, and « non-grammatical » to a randomly built item, as if they had learned some of the rules of the grammar used to generate the items seen during the learning phase.

The subjects are unable to say anything about the rules.


Step 1: The circles light up one by one in immediate succession, forming a series of 5 events (e.g., the series : B C B A C ) Step 2: The subject has to predict which will be the 6th lighted circle (A, B or C ?) . Step 3: The correct response is provided in feedback Step 4: Next trial (step 1).

A

CB

Task of Kushner, Cleeremans & Reber (1991)

????

correct

Positions 1, 3 and 5: random. Positions 2 and 4: random, excluding repetition Position of the 6th circle depends on positions 2&4.

A

CB

B C B A C C

Results : 33% (random guess) at beginning ↦ ↦ ↦ 45% (> random) at the endSubjects are unable to verbalize anything about the rules.

➠ if 2&4 clockwise, then 6th = B➠ if 2&4 counterclockwise, then 6th=C➠ if 2=4, then 6th = A


3. A developmental approach

Implicit Learning and DevelopmentImplicit Learning and Development

Implicit learning processes are surely at the origin of various new behavioral acquisitions in the course of development

few experimental studies, but a strong postulate (Reber):

independency of these processes in relation to age.

some studies with babies (experiments on language or on vision) demonstrate the early efficacy of these processes;

some studies with children, mainly between 6 and 11 years (6 with SRT, 2 avec covariations, 3 avec AG).

a strong tendency to confirm Reber’s postulate.

Our own experimental paradigm Our own experimental paradigm

We aimed at studying the action of unconscious influences on learning, using the same task, through a large range of ages and in children with mental retardation.

We selected a drawing task of geometrical figures.

This task is naturally structured by a regularity : the « start-rotation principle (SRP)» described by Van Sommers, 1984.our idea was to manipulate this natural regularity without the subject knowing it.

The « Start-Rotation Principle »The « Start-Rotation Principle »

It structures the graphic production of closed geometrical figures.

covariation between the starting point and the movement’s direction

around 70% of our spontaneous production.

it escapes to a strong decisional control for most subjects

Illustration for circle drawing:

11

5

The performed experimental The performed experimental manipulation:manipulation:

scope: through a repeated practice, we tried to lead the subjects to reverse the principle, without them knowing it : i.e. we tried to lead them to learn to associate a clockwise rotation with a top starting point on the one hand, and a counterclockwise rotation with a bottom starting point on the other hand.

Task for the subjects: We said them that we wanted to study their degree of expertise in graphic activity, asking them to be both fast and accurate when they traced over geometrical figures.

The drawing taskThe drawing task

We designed 3 groups of learning: - Learning with a SRP at 80%- Learning with a SRP at 20% : group of interest- Control group (free tracing over figures)TEST: identical for all subjects, tracing with

only the starting locations imposed

We told them that we introduced variability in their movements by imposing different starting locations and movement directions.

The « rational » of the test (after The « rational » of the test (after learning):learning):

If the practice during the learning phase has an impact, the subjects should, in test, behave differently as a function of their initial learning group: their respect for the SRP at test should reflect their learning condition.

consequently, the group trained at 20% should respect the SRP significantly less than the control group, or than the group trained at 80%.

This task is an implicit learning task:This task is an implicit learning task:

Neutrality of the instructions:The subject’s attention was oriented on the requested accuracy and speed of tracing, and he was ignorant with respect to the regularity present in the task (SRP at 80% or 20%).

Neutrality of the measured behavior:The behavior that the experimenter tried to modify without the subject knowing it (reversing the SRP) was different from the behavior that the subject tried to perform at a conscious level (being good at tracing)

Only unconscious processes may lead subjects to modify their spontaneous practice.

Our experiments using this paradigm :Our experiments using this paradigm : Vinter & Perruchet, 2000:

Testing if the SRP can be incidentally reversed through appropriate practice, testing the age-independency postulate and whether the resulting implicit learning is maintained in time.

Vinter & Perruchet, 2003 : Testing if the motor component play a major role in the type of implicit learning involved in our tracing task.

Vinter & Detable, 2003 : Testing the IQ-independency postulate (Reber), comparing typically developing children with intellectually impaired children.

Detable & Vinter, 2004, 2006 : Testing the maintain, in time, of an implicit learning in intellectually disabled children, and the possibility of transfert of this learning in a similar task.

Vinter & Detable, in revision : Comparing the effects of an explicit and implicit learning in children with or without intellectual retardation.

Immediate Test: 4-10 years and adultsImmediate Test: 4-10 years and adults

0

10

20

30

40

50

60

70

80

4 5 6 7 8 9 10 ad

% S

RP

free

80%-SRP

20%-SRP

A question….A question….

Do these implicit learning processes keep their efficacy in time ?

Introducing a delay between the learning phase and the testing phase

Differed test: children 4-10 yrs and adults.Differed test: children 4-10 yrs and adults.

010

2030

4050

6070

8090

4 5 6 8 10 ad

% S

RP

free

80%-SRP

20%-SRP








Would the behavioural modifications induced in our task be elicited even if we suppress the motor component, asking subjects to look at figures being traced on a computer screen?

Subjects showed figures traced on the screen, applying the SRP either at 20% or at 80%.

Perceptual learning: 6-10 years and Perceptual learning: 6-10 years and adults.adults.

0

10

20

30

40

50

60

70

6 8 10 ad

% S

RP

80%-SRP

20%-SRP








We slightly modified our tracing task so that low-IQ children could be studied:- only a training at 20% with pre-test, training and post-test.- only circles.

Children/Adolescents with mental Children/Adolescents with mental retardationretardation

0

10

20

30

40

50

60

70

80

90

SRP

Enf/léger Enf/moyen Ado/léger Ado/moyen

AvantAprèsB

DiscussionDiscussion

Efficacy Efficacy of these implicit learning processes of these implicit learning processes andand independency independency from age and IQ.from age and IQ.

These processes lead always to modify These processes lead always to modify behavior such that it reflects the behavior such that it reflects the structural structural characteristicscharacteristics or or regularitiesregularities of a situation of a situation, , without the subject knowing it.without the subject knowing it.

The resistance of this phenomenon through The resistance of this phenomenon through age does not involve age does not involve that no differences between ages can be shown. Learning implicitly is surely possible at all ages during development, from infancy to the old age, but according to the complexity of the situations, it also may be more or less easy to elicit, more or less efficace.


4. About an experiment currently in progress…

Our current direction of researchOur current direction of research

Our tracing task is a simple task, structured by a Our tracing task is a simple task, structured by a single covariation. This simplicity could explain single covariation. This simplicity could explain why we did not observe any age-related why we did not observe any age-related differences in our experiments.differences in our experiments.

we use the artificial grammar we use the artificial grammar paradigm, with colors as elements (instead of paradigm, with colors as elements (instead of letters). The grammar withe colors is used to letters). The grammar withe colors is used to build flags.build flags.

We are also trying to test specific hypotheses We are also trying to test specific hypotheses coming from different theoretical models of coming from different theoretical models of implicit learning.implicit learning.

Artificial grammarArtificial grammar

with colors

The models of implicit learningThe models of implicit learning Abstraction of rulesAbstraction of rules (Reber, 1976): (Reber, 1976):

existence of a highly sophisticated unconscious existence of a highly sophisticated unconscious processor devoted to rules’ abstraction.processor devoted to rules’ abstraction.

Learning exemplarsLearning exemplars (Brooks, 1978): (Brooks, 1978): Subjects would memorize the items presented Subjects would memorize the items presented during the learning phase and would make during the learning phase and would make grammaticality judgments on the basis of the grammaticality judgments on the basis of the similarity between the items seen during learning similarity between the items seen during learning and the items seen at test.and the items seen at test.

Learning fragmentsLearning fragments (Perruchet, 1994): (Perruchet, 1994): Subjects would memorize fragments of items, Subjects would memorize fragments of items, because attentional processes imply a natural because attentional processes imply a natural segmentation of the presented informationsegmentation of the presented information..

Formation of cognitive units : SOCFormation of cognitive units : SOC (Perruchet&Vinter, (Perruchet&Vinter, 2002):2002): Subjects build cognitive units from associations Subjects build cognitive units from associations between primitives that enter together in the between primitives that enter together in the conscious focus.conscious focus.

Artificial grammarArtificial grammar

with colors

A study in progress (PhD of A. Witt)A study in progress (PhD of A. Witt)

Illustration of the artificial grammar used (a grammar that makes possible the generation of salient regularities: doublets and symmetries).

Grammatical items : frequent doublets Grammatical items:

frequent symmetries

Series of random items

Exposition to the flags during the learning Exposition to the flags during the learning phasephase..

Testing phaseTesting phase

Hypotheses ?Hypotheses ?

1. Expected differences between the 3 groups:

- can we expect that the group confronted with the “doublets’ set” produces more grammatical doublets in test than the 2 other groups, and that the group confronted with the “symmetries’ set” produces more grammatical symmetries in test than the other groups ?

Differences between groups ?Differences between groups ?

The The abstractionnistabstractionnist positionposition predicts an absence of differences between predicts an absence of differences between the « doublets » and the « symmetries » groups: the the « doublets » and the « symmetries » groups: the grammar is identical, all the rules are present in the grammar is identical, all the rules are present in the items observed during learning. A relative items observed during learning. A relative independency exists between the impact of learning independency exists between the impact of learning and the specific set of grammatical items that are and the specific set of grammatical items that are actually shown to the subjects.actually shown to the subjects.The other positionsThe other positions they all predict a relationship between the they all predict a relationship between the items shown to subjects and the learning effect: items shown to subjects and the learning effect: memorization of the seen exemplars, of fragments memorization of the seen exemplars, of fragments or of salient units from these exemplarsor of salient units from these exemplars. . We consequently expect a production of We consequently expect a production of grammaticalgrammatical doublets superior in the doublets doublets superior in the doublets group than in the symmetries group or in the group than in the symmetries group or in the control group (and conversely, of symmetries in the control group (and conversely, of symmetries in the symmetries group).symmetries group).

Difference between groups Difference between groups Symmetry/DoubletsSymmetry/Doublets

01234567

%

Symmetry Doublet Control

Production of gramm. doublets

Difference between groups Difference between groups Symmetry/DoubletsSymmetry/Doublets

0

1

2

3

4

5

6

%


Production of gramm. symmetries

Hypotheses ?Hypotheses ?

2. Differentiation between the exemplarist, fragmentarist and SOC models:

- can we expect that the group confronted with the “doublets’ set” develops a sensibility to the very “doublet” regularity (regardless of the colours) and the group “symmetry” a sensibility to the very “symmetry” regularity ?

Sensibility to repetition/symmetry ?Sensibility to repetition/symmetry ?

SOC PositionSOC Position The conscious experience of seeing yellow-red-The conscious experience of seeing yellow-red-yellow or blue-blue concerns specific color yellow or blue-blue concerns specific color successions but also contains a consciousness of the successions but also contains a consciousness of the very repetition structurevery repetition structure. The direct encoding of . The direct encoding of these relational structures (e.g., XYX or XX) may these relational structures (e.g., XYX or XX) may progressively lead to a sensibility to symmetry for progressively lead to a sensibility to symmetry for itself or to repetition for itself.itself or to repetition for itself. Exemplarist/fragmentarist positionsExemplarist/fragmentarist positions they do not predict a sensibility to the they do not predict a sensibility to the symmetry or repetition structures because only the symmetry or repetition structures because only the items or fragments of items that have been items or fragments of items that have been seenseen are are memorised (only specific successions of colors are memorised (only specific successions of colors are encoded). encoded).

Sensibility to repetition/symmetrySensibility to repetition/symmetry

0

1

2

3

4

5

6

%


Production of (nonG) doublets

Sensibility to repetition/symmetrySensibility to repetition/symmetry

0

2

4

6

8

10

%


Production of (nonG) symmetries

Which model ?Which model ?

Other investigations have to be made before concluding.

What seems clear is that models based on memory (such as the exemplarist, fragmentarist and SOC models) are more plausible than a model based on unconscious abstraction

Conclusion: a useful learning mode ?Conclusion: a useful learning mode ?

This ongoing experiment shows that This ongoing experiment shows that children can learn implicitly relational children can learn implicitly relational structures, ie they can incidentally structures, ie they can incidentally become sensitive to relatively become sensitive to relatively abstract relations:abstract relations:

to what extent can this implicit to what extent can this implicit mode of learning be used to mode of learning be used to remediate learning difficulties ?remediate learning difficulties ?

to what extent is it possible to to what extent is it possible to design implicit learning situations? design implicit learning situations? What are the characteristics of any What are the characteristics of any implicit learning situation?implicit learning situation?

Characteristics of the learning Characteristics of the learning situations situations

No exposition to No exposition to «« errors »errors »: : confrontation solely to positive confrontation solely to positive instances of the rules or regularities instances of the rules or regularities (during learning).(during learning).

implicit learning mechanisms implicit learning mechanisms capture frequentist (statistical) capture frequentist (statistical) information, regardless of its origin information, regardless of its origin (from positive instances or errors)(from positive instances or errors). . We can learn implicitly as well We can learn implicitly as well correct as incorrect information: we correct as incorrect information: we learn what is frequently presentedlearn what is frequently presented..


Simplify the situations and isolate at Simplify the situations and isolate at best the regularities of interest:best the regularities of interest:

the attentional focus is limited the attentional focus is limited and can process only a reduced number and can process only a reduced number of elements at the same time;of elements at the same time;

it is also constrained in time and it is also constrained in time and space: informations that have to be space: informations that have to be associated must not be distant in time or associated must not be distant in time or in space.in space.


Lead the subject to process the desired Lead the subject to process the desired information without any explicitation information without any explicitation (e.g.: spelling out and copying words in (e.g.: spelling out and copying words in order to make the subject aware of order to make the subject aware of correct orthographical instances).correct orthographical instances).

Finally, necessity to repeat several Finally, necessity to repeat several times the learning situation (implicit times the learning situation (implicit learning needs time to occur).learning needs time to occur).

Conclusion: a useful learning mode ?Conclusion: a useful learning mode ?

WarningWarning: the implicit learning mode : the implicit learning mode cannot simply substitute the explicit cannot simply substitute the explicit learning mode in most educational learning mode in most educational contexts:contexts:

itit is not appropriate if one wants is not appropriate if one wants that children acquire explicit knowledge.that children acquire explicit knowledge.

A possible association : the implicit mode A possible association : the implicit mode can make subjects familiar to the rules (to can make subjects familiar to the rules (to their products), and this familiarity may their products), and this familiarity may facilitate the further action of explicit facilitate the further action of explicit learning. learning.

Thanks for your attention!

Implicit Learning processes in Children

Documents

explicit learning

learning condition

implicit learning processes

duration of learning

type of learning

conscious implicit learning

music implicit learning

useful learning mode