Visuo- spatiaalinen - Jyväskylän yliopistousers.jyu.fi/~prasanen/Files/Visuospatiaaliset taidot.pdf · Paikkasolut Paikkasolut prosessoivat spatiaalista muistia siitä, missä jokin

Visuo-spatiaalinen hahmotus

NMI- PR

Vertailevan tutkimuksen menetelmistä

Rat Radial Maze (Olton)(from Goodlett et al, 1992)

Human Radial Maze (from Overman et al., Beh.

Neurosci., 110:1205-1228, 1996)

Vertailevan tutkimuksen menetelmistä

Rat Morris Maze(from Goodlett & Johnson,1999)

Human Radial Maze (from Overman et al., Beh.

Neurosci., 110:1205-1228, 1996)

PaikkasolutPaikkasolut prosessoivat spatiaalista muistia siitä, missä jokin on

Korostunut esimerkiksi linnuilla, jotka säilövät ruokaa

Hippokampuksessa

4

Hippokampus ja taksikuskit

Lontoon taksikuskeilla on takanaan massiivinen harjoittelu siitä, missä mikin paikka on

Aivokuvantamisella todettu, että taksikuskien hippokampus on suurempi kuin normaaleilla kontrolleilla

(Maguire & al., 2000)

5

TEAR DUCTANDDRAINAGECANAL

PUPIL IRISSCLERA

Silmä

SilmäValoreseptorit

Sauvasoluthämäränäkö ja liikkeen havaitseminensensitiivisiä laajalle valospektrilleeri erota värejäharmaan sävyjen erottelu

Tappisolutvärien erotteluun

8

SilmäRetinan keskellä tappeja

tarkkaavaisuuden kohteena olevan objektin yksityiskohtainen havaitseminen

Reunoilla sauvoja

liikkeen havaitseminen perifeerisessä näössä

9

Värit“Fotopigmentit” aistivat valoa

jokainen sensitiivinen eri aallonpituudelle

kolmea eri tyyppiäsininenvihreäpunainen (keltainen)Muut värit niiden yhdistelmiä

10

VärispektriTrichromatic (Young-Helmholtz) Theory:

Väriteri määrä tappisoluja

pääosin punaisia (64%), vähän sinisiä (4%)

huonompi sensitiivisyys lyhyille aallonpituuksille

retinan keskellä ei sinisiä tappejapienet siniset objektit katoaa fiksaatiossa

12

Sokea piste

Primaari visuaalinen kortex (V1)

Yksikertaiset solut = reagoi tietyssä paikassa tiettyyn orientaatioon

Kompleksit sovlut = reagoivat tietyssä orientaatiossa olevaan tietynsuuntaiseen liikkeeseen

Hyperkompleksit solut = reagoivat tiettyihin kombinaatioihin (koko, muoto, orientaatio, suunta ja liike)

14

Havainnon muodostuminen koostuu useista toisistaan riippumattomista prosesseista. Pään sisällä ei ole “katselijaa”.

Valikoiva tarkkaavaisuus

XX

O XX

OXX

X

X

XX

XX

Set Size

RT

Feature Search: Find O

XOO

XXOOO

O

O

O X

XX

Set Size

RT

Conjunction Search: Find O

Visuaalinen etsintäFeature Integration Theory (Treisman)

kaksivaiheteoriapreattentiivinen vaihe rekisteröi peruspiirteet kuten orientaatio ja väriattentiivinen vaihe yhdistää piirteet yhteen “objekteiksi ja vaatii tarkkaavaisuutta

teorian ongelma: myös kompleksi kohde voi “hypätä esiin”

17

Visuaalinen etsintä

A3 J

E

N

G

XF

MKH C

Etsi 3 (eri kategoria)

A

B

JEN

G

X F

MKHC

Etsi B (sama kategoria)

Merkityksen vaikutus



Piirre, johon kiinnitetään huomiota säätelee siihen piirteeseen liittyvien aivoalueiden neuraalista aktiviteettiä

20

Näkeminen ja havaitseminen

NäköhäiriötSokeus on suhteellinen käsite, jolla tarkoitetaan erittäin voimakasta näön tarkkuuden heikkotta tai näkökentän kapeutta (alle 20°) molemmissa silmissä

synnynnäisen sokeuden esiintyvyys länsimaissa 1-8/10 000 syntynyttä lasta kohden

myöhempiä sokeutumisia noin 25% synnynnäisten määrästä

22

Visuaaliset häiriötSynnynnäisistä visuaalisista häiriöistä kärsivillä usein muitakin vammoja

17% kuulovaurioita46% epileptisiä kohtauksia78% kehitysviivästymiä (Gradner, 1986; Wilkinson, 1987)

Tyypillisiä aiheuttajiaalhainen syntymäpainoperinnölliset häiriötraskaudenaikaiset infektiotvaikea/avustettu synnytys

23

Silmän liikkeetKarsastus eli strabismus voi johtaa

stereoskooppisen näön puutteeseenkaksoiskuviin (lapsilla harvinaisia, koska lapset pyrkivät häivyttämään toisen kuvan välttääkseen kaksoiskuvaa)

Silmänliikkeiden häiriöt voivat johtuasilmän lihaksiston häiriöistä tai keskushermostoperäisistä syistä, kuten

vauriot (hapenpuute, infektiot)aivohermojen puristus (kallonsisäisen paineen kasvu)myrkytystilat

24

Muita näköhäiriöitä

Ambylopiapysyvä tai hetkellinen näön tarkkuuden heikkeneminen ilman vauriota retinassa tai näköhermoradoissa

voi ilmetä sekundaarisena verenkiertohäiriöissä tai kasvaimen aiheuttamissa puristustiloissayleisin syy sensorinen deprivaatio varhaislapsuudessalapsilla havaittu usein keskitasoa heikompi ÄO (Stewart-Brown ym. 1985)

Akromatopsia eli värien erotteluvaikeusProsopagnosia eli kasvojen tunnistusvaikeus (voi olla visuaalinen)

25

Takaraivolohkon vauriot

Riippuen vaurioiden laajuudesta seurauksena voi olla havaintohäiriöt “sokeista pisteistä” kokonaiseen kortikaaliseen sokeuteen

Lapsilla toipuminen huomattavasti suurempaa kuin aikuisilla

usein jäännösoireina korkeamman tasoisia havaintoprosessoinnin häiriöitä

26

Kortikaaliset näköpuutteet (V1)

Skotooma

Quadrantanopia

Hemianopia

Näkökenttäpuutokset

Visuaalinen ja spatiaalinen havainto

Erilaiset solut, eri kerrokset

Syöttö muiltaaivoalueilta

Magno syöttö muilleaivoalueille:

(3) liike V5:lle

Parvo syöttö muilleaivoalueille:

(1) muoto V3:lle, &(2) väri V4:lle

Makaki-apinan näköaivokuoren kytkentäkaavio

Tai yksinkertaisemmin

V1-V2 V3 V4 V5

Muoto

Väri

Liike:global& local

Kaksi reittiä

Mikä

Missä, miten

Mitä -reitti “Ventraalinen” reitti temporaalilohkolle

yhteys verbaaliseen“tietoinen” visuaalinen havaitseminenobjektien tunnistaminenmuistiallosentrinen: tuttujen suhteiden ja tilojen navigointi (kognitiivinen kartta)soluja, jotka reagoivat tiettyihin kohteisiin (esim. kädet, kasvot)

34

Missä -reitti “Dorsaalinen” reitti parietaalilohkolle“tyhmä”, ei yhteyttä verbaaliseenehkä paremminkin “Miten” -reittimagnosellulaarisolu-yhteys myös superior colliculuksestaegosentrinen navigointi tilassasaattaa sisältää peilisoluja, jotka reagoivat toisen toimintaan

35

Kuinka havaitaan syvyys?

2 silmää

monokulaarinen vihje 1

Takana oleva on kauempana

Interpositio


Koon tuttuus: aiempi tieto

koosta

monokulaarinen vihje 3-1

Lineaarinen perspektiivi: Mitä kauempana, sitä pienempi

monokulaarinen vihje 3-2

Jos kaksi objektia projektoituvat retinalle samansuuruisina, kauempana oleva nähdään suurempana.


Atmosfäärinen perspektiivi:Mitä kauempana, sitä epätarkempi(ja mitä enemmän roinaa välissä, sitä kauempana)


Liikeparallaksi:liikkuvan ohittaessa liikkumattomat, lähempänä olevan siirtyvät nopeammin


Muotovihjeet


Ohut linssi - kaukana

Paksu linssi - lähellä

binokulaarinen vihje 1

45° 20°

Konvergenssi: kulma kasvaa lähestyttäessä

binokulaarinen vihje 2

Retinaalinen eroavaisuus:Molemmilla silmillä on hieman eri näkökulma...

Visuaaliset prosessit

Yleisin jako kolmeen osa-alueeseen(esim. Chalfant & Scheffelin, 1969)

Spatiaalinen orientaatio (parietaali)Visuaalinen erottelu (okkipito-temp)Havaitun tunnistaminen (temporaali)

ei kovin suurta yksimielisyyttä

48

Spatiaaliset prosessit

Ei edelleenkään yksimielisyyttäLinn & Petersen (1985)

Spatiaalinen havaitseminenkyky mieltää avaruudellisia suhteita

Mentaalinen rotaatiokyky kääntää ja pyöritellä mielessään 2- ja 3-ulotteisia kuvia tai kuvioita

Spatiaalinen visualisaatiokyky käsitellä spatiaalista informaatiota useissa vaiheissa ratkaisuun pääsemiseksi

49

Sukupuolierot

Card Rotations (.31)

3-D Mental Rotations (.67)

Embedded Figures (.18)

Paper Folding (.12)

Meta Analysis (N = 286) (Voyer, Voyer & Bryden, 1995)

Havainnon oppimisen kaksi mekanismia

Bottom-up: adaptaatio

toistuvan ärsytyksen aikaansaama hermoverkon automaattinen muovautuminen

Top-down: perseptuaalinen oppiminen

ylempien aivorakenteiden ohjaama alempien aivorakenteiden asteittainen erikoistuminen

51

Reverse Hierarchy Theory

signal-to-noise ratio is poor, particularly with respect tospatial aspects, because generalized object selectivity isobtained in part by convergence across spatial parameters.

Learning is therefore attention driven, where attentionis the mechanism for choosing the relevant neuronalpopulation, by increasing its functional weight. It followsthat initial high-level learning must precede low-levellearning, as it provides the essential enabling stage for thebackward search process.

The Eureka effectAn extreme condition occurs when subjects are trainedonly with difficult cases, so that they have no introductoryeasy trials along which to learn. Under these conditions,typically no learning occurs. However, a single exposure tothe stimulus (as it appears on the screen) suffices toinitiate an immediate learning process [21,29]. We termedthe impact of this single exposure, which enables subse-quent learning the Eureka effect (see Box 1). In terms offormal definitions, Eureka is a special case of priming. Butwhether such cases are indeed manifestations of the sameunderlying mechanisms is still unclear.

Theoretical and experimental challenges to RHTThe psycho-anatomy logic assumes a direct relationshipbetween typical receptive field properties and the area’sgeneral function. This concept has been questioned [30] ongrounds of cortical variability, claiming that althoughaverage receptive field size and orientation tuningbroaden along the cortical hierarchy, substantial varia-bility at any stage produces significant overlap betweenareas, including presence of small receptive fields athigher levels. Thus, even specific learning could stem fromhigh-level modifications (discussed in [31]). This criticismin fact relates to the field’s lack of knowledge regarding theexact relations between single receptive field propertiesand the area’s computations (i.e. the neural code).

Currently, psycho-anatomical assumptions go beyondwhat has been directly substantiated. Still, the assump-tion that average receptive field properties denote thearea’s functional resolution is sensible, particularly whenone assumes a population code. Moreover, the majority offindings in the visual learning literature are naturallyaccounted for by the RHT dynamic view.

An important RHT prediction is that perceptuallearning will be contingent on task-specific attention,because the backward search for increased signal-to-noiseratio is attention-driven. Many studies, using ‘pop-out’detection [17,32], orientation and texture discrimination[15,33], and Vernier acuity [34,35], are consistent withthis prediction. However, recent studies in the visual[36,37] and tactile [38,39] modalities found that passivestimulation can also improve discrimination abilities.

In the visual modality, Watanabe et al. [36] found thatwhen observers performed a difficult visual task nearfixation while also exposed to surrounding moving dots,subsequent discrimination between the motion directionto which observers were continuously exposed and othermotion directions was significantly improved. Thus,observers improved in motion discrimination even thoughthe motion signal was not attended and had a sub-threshold coherence level (5%). Namely, it had not evenbeen consciously perceived. Initial improvement was morespecific to the local direction of motion and to its retinalposition than subsequent improvement [40], suggesting abottom-up dynamics of modifications. Both character-istics, namely bottom-up dynamics and no need forattentional control, have the flavor of adaptation pro-cesses (e.g. [41,42]). Such processes can lead to increasedsensitivity to changes around the massively exposedstimuli (improved novelty detection). Still, the relationsbetween this form of plasticity and practice-inducedperceptual learning are not well understood.

It should be noted that RHT does not assert that thereare no bottom-up induced modifications, such as adaptation

TRENDS in Cognitive Sciences

stimulus

…

Reverse Hierarchy Theory:visual system hierarchies of areas and cell types

Explic

it feedback

connecti

ons

Implic

it feedfo

rward

connecti

ons

Figure 4. Reverse Hierarchy Theory. Initial vision at a glance depends on high-levelobject and category representations built by implicit hierarchical processing [62]. Inthis way, initial high-level learning transfers over basic stimulus parameters [21].Later vision with scrutiny is a return to simple feature details available at low levels[62]. Thus, later low-level learning is parameter-specific, being a low-levelmodification by guided return down the Reverse Hierarchy [21].

Box 2. Transfer along a continuum

As found in our studies, training is more effective if subjects startwith easy conditions and gradually move to more difficult con-ditions. The importance of beginning training with easy conditionswas first found by Pavlov. When Pavlov reinforced a dog’s salivationfollowing its seeing an ellipse but not following its seeing a circle(Figure I, rightmost pair of stimuli), the dog could not avoidgeneralization and salivated at sight of the circle, as well. Only byusing very elongated ellipses, and training along the continuum,from left to right, was it able to achieve good performance for smallcircle/ellipse differences. This phenomenon was subsequentlytermed ‘transfer along a continuum’ (of different degrees ofdifficulty).


intermediatehardest

smallest

easiest

largest

Difficulty:

Difference:

Figure I.

Opinion TRENDS in Cognitive Sciences Vol.8 No.10 October 2004460

www.sciencedirect.com

RHT: Visuaalisen oppimisen vaiheet

progressively established which integrates task-relevantselected information at each level.

(4) Expert performers, who have had substantialamounts of training, can base their performance on higherlevels again, even in difficult cases. As described above(highly-trained performers) there is a gradual process ofmodifications at lower and lower levels. Lower levelmodifications affect not only the levels of modificationthemselves, but also all higher levels that receive inputfrom these levels. Consequently, higher-level represen-tations are gradually better and better tuned for perform-ance of the task as they attain an improved task-relatedsignal-to-noise ratio. As there is always a preference foraccessing the highest level that has sufficient task-relatedsignal, higher-levels will be the basis for performanceagain. These levels have again the advantage of beingmore broadly tuned, because they are the outcome ofconvergence of multiple components. The increasedweightings of informative inputs – at all cortical levels –now bias higher-levels towards the trained domain, whichgains dominance compared with its emphasis in theoriginal naı̈ve distribution. A schematic illustration ofthis concept is shown in Figure 5.

This process yields the special characteristics of experts’perception. It is immediate and holistic on the one hand,suggesting high-level representations. Yet the generaliz-ation of their expertise to untrained conditions is limited,consistent with lower-level plasticity [60]. According toRHT, experts access high-levels again, but their high-levels are now biased towards the trained stimulusdomain owing to the earlier top-down cascade ofmodifications.

The holistic nature of expert performance is capturedin the classical term ‘chunking’ – referring to a processduring which separate components gradually become asingle perceptual or cognitive entity. A classic example isthat of Morse decoders who gradually hear whole wordsrather than single bits. A timely example is that of expertvideo-game players who develop a spatially holistic modeof performance [61]. For example, they can enumerate alarger number of spatially distributed elements withoutcounting. Interestingly, improved performance fromvideo-game training occurred not only in the spatialdomain but also in the temporal domain: video-gameplayers had far less of an attentional blink than non video-game players, outperforming them even for lag one.

RHT attributes chunking and effortless performance toexperts’ ability to rely on high-level representations again.Even so, to explain the breadth of attentional changescharacterizing expert performance [61], it should befurther extended (see also Box 3).

RHT and perceptionThis article has focused on RHT and learning. Howeverlearning is not an odd case of perception. Rather it reflectsthe sequence of perception, attention and retention.Hence, RHT is expected to apply to perception in general.Specifically, learning begins at high-levels because theseare the first levels accessed by conscious perception.Consequently, with brief exposures, we consciously per-ceive the ‘gist of a scene’, but not its fine details. Perceivingdetails requires access to lower levels, therefore more timeand scrutiny [62]. We here summarize RHT’s predictionsfor search tasks,whichwerethe focusof our learningstudies.


Simple

. . . . . .

IT

Perceptual learning stages

Complex

IT

original detector diversitylow signal-to-noise

enhanced detectorsin expert system

(a) (b)

anterior IT

Figure 5. From naı̈ve to trained to expert performance. (a)Naı̈ve performance is based on activity at high-level cortical representations (which are formed by fast feedforwardprocessing; arrows at left). This representation suffices for normal daily activity when scenes are simple and meet expectations. For difficult conditions and unexpectedcircumstances, conscious vision must seek details of the scene in lower areas, depending on top-down guidance over activated pathways. Training for these tasks enablessubjects to perform this backward search more efficiently and to use the most appropriate level for the task on a trial-by-trial basis (dotted downward arrows). (b) Highlytrained subjects have modified receptive fields first at higher and then also at lower cortical levels. Experts are those whose higher level representations have been modifiedby adding weight to appropriate inputs and pruning uninformative inputs (for the trained task). Performance can thus once again depend on higher, more generalizing, units.



EUREKA- ilmiö

Ahissar, M., & Hochstein, S. (1997). Task difficulty and specifity of perceptual learning. Nature, 387, 401-406.

The reverse hierarchy theory of visualperceptual learningMerav Ahissar1 and Shaul Hochstein2

1Department of Psychology and Interdisciplinary Center for Neural Computation, Hebrew University, Jerusalem, 91905, Israel2Department of Neurobiology and Interdisciplinary Center for Neural Computation, Hebrew University, Jerusalem, 91905, Israel

Perceptual learning can be defined as practice-inducedimprovement in the ability to perform specific percep-tual tasks. We previously proposed the Reverse Hier-archy Theory as a unifying concept that links behavioralfindings of visual learning with physiological andanatomical data. Essentially, it asserts that learning isa top-down guided process, which begins at high-levelareas of the visual system, and when these do notsuffice, progresses backwards to the input levels, whichhave a better signal-to-noise ratio. This simple concepthas proved powerful in explaining a broad range offindings, including seemingly contradicting data. Wenow extend this concept to describe the dynamics ofskill acquisition and interpret recent behavioral andelectrophysiological findings.

Throughout life our sensory receptors are continuouslybombarded by stimuli. This activation not only inducesperception, it also modifies our representation mechan-isms, thereby affecting all subsequent perception. Recentevidence shows that a large degree of perceptual plasticityis retained in adulthood, with long-term manifestationsincluding adaptation, priming and perceptual learning.Currently, these different phenomena are defined by theirbehavioral characteristics and the manner in which theyare induced, rather than by their respective underlyingneural mechanisms.

This article focuses on perceptual learning, defined aspractice-induced improvement in the ability to performspecific perceptual tasks (see [1] for a classic review and[2] for recent overviews). We shall argue that whattypically limits naı̈ve performance is the accessibility oftask-relevant information rather than the absence of suchinformation within neuronal representations [3]. We shallpresent the reverse hierarchy theory (RHT) of perceptuallearning asserting (i) that perceptual improvement lar-gely stems from a gradual top-down-guided increase inusability of first high- then lower-level task-relevantinformation, and (ii) that this process is subserved by acascade of top-to-bottom level modifications that enhancetask-relevant, and prune irrelevant, information (see [4]for a computational model applying a similar concept).

The relations between plasticity processes duringsubstantial practice and those dominating the first fewexposures, are not well understood. Only the first are

typically referred to as perceptual learning, whereas thelatter are termed priming. Recent evidence suggests that,at least when governed by top-down control, singleexposures (priming) can induce strong and long-lastingeffects that clearly change our perception (see Box 1),suggesting that these might be the initial processes ofperceptual learning, as described below (TheEureka effect).

Box 1. The Eureka effect

The picture (Figure I) generally appears simply as a set of gray andblack regions, without further meaning; the object represented ishard to categorize without further cues. Following considerableinspection, (or a single look at the ‘clue’ in Figure 1 overleaf), thepuzzle is solved, and, without further practice, observers directlyperceive a bearded figure. This effect is rapid, strong and long-lasting, suggesting that significant top-down control determines ourconscious perception. Ahissar and Hochstein [21] found that a singlelong exposure to a ‘pop-out’ stimulus enabled learning of a verydifficult detection task, based on brief and strongly maskedpresentation of similar stimuli – a task that was almost never learnedwithout the Eureka enabling experience. Thus, it appears that similartop-down control or guidance mechanisms influence both percep-tual learning and conscious perception. An important differencebetween these effects is that in the experiment, following the singleeasy-case ‘Eureka’ exposure, hard-case perceptual learning wasenabled, but still required; no such training is needed for Figure Ihere.

Figure I.Corresponding author: Merav Ahissar ([email protected]).

www.sciencedirect.com 1364-6613/$ - see front matter Q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2004.08.011


vihje

A third phenomenon of perceptual plasticity is adap-tation. This seems to be a more basic mechanism that isnot specific to the nervous system and is characteristic ofall biological systems. It differs from perceptual learningin being induced by exposure to stimuli rather than bytask-specific practice. Thus, it is essentially a bottom-upprocess, in which internal representations are modified inresponse to the current distribution of external stimuli[5,6]. Consequently, responses to unvarying stimuli arereduced, inducing increased sensitivity to changes, ornovelty detection. Because perceptual training involvesexposure to stimuli and consequently adaptation, it isdifficult to study perceptual learning without inducingadaptation processes, whose impact on task performanceis hard to discern (see discussion in [7] and an experi-mental example of bias induced by adaptation in [8]).

Psycho-anatomy logic and the Reverse Hierarchy TheoryThe term psycho-anatomy, coined by Julesz more than 30years ago [9], implies that we can deduce from behavioralfindings information regarding the underlying anatomicalstructures. In the visual modality, our knowledge of basicrepresentations is relatively broad, basedmainly on singleunit receptive field characteristics, and recently oncorroborations from fMRI studies, [10–12]. Particularlywell understood is the representation of oriented light ordark bars and edges. In the primary visual cortex, V1,single neurons are selective for orientation and retinalposition with relatively narrow tuning curves in thesedomains (the ‘Simple’ cells of Hubel andWiesel, [13]). Bothspatial and orientation tuning curves are broader forneurons at higher visual areas along the visual hierarchy.

This basic observation can be used to deduce the site ofneuronal modifications from learning generalization.Namely, if we train subjects on a perceptual task using a

set of stimuli whose orientation and retinal position arefixed, then changing the position or orientation of thestimulus would lead to activation of a non-overlappingpopulation of neurons at lower-level areas. Thus, if themajor bulk of plasticity underlying behavioral improve-ment occurred at low-levels, improvement would nottransfer to these new stimulus conditions and subjectperformance would be degraded towards initial levels,requiring a process of re-learning. On the other hand, iflearning resulted from high-level modifications, it wouldlargely transfer to novel positions and orientations. Thesepossibilities are schematically illustrated in Figure 2 forSimple and Complex V1 neurons [13], and higher-areaneurons (inferotemporal cortex, IT; [14]).

This psycho-anatomy logic has been used in manypsychophysical studies to deduce the underlying site ofplasticity from the degree of learning specificity to spatialdimensions such as the trained eye [15], retinal positionand orientation (e.g. [16–19]). However, parallel studiesproduced contradictory findings regarding learning speci-ficity even when experiments seemed rather similar. Forexample, Karni and Sagi [15] found remarkable speci-ficity, including to the trained eye, whereas Schoups et al.[20], using a very similar texture discrimination task,found complete generalization across eyes.

We decided to study directly whether this variabilityitself follows systematic rules. The task we used wasdetecting the presence of an oddly oriented bar in an arrayof homogenously oriented distractor bars [17–19,21,22], asillustrated in Figure 3. Inmost of the stimuli that we used,the orientation of the target greatly deviated from that ofthe distracting bars, yielding effortless detection,accompanied by the notion that the odd element ‘popsout’ ([23]; see review in [24]). That is, with long exposures(O250 ms), this task is trivial to begin with, andperformance (reaction time) is independent of the numberof distractor elements in the array. We made it difficult byusing brief exposures followed by a masking stimulus(Figure 3). Under these conditions, task difficulty could becontrolled in several ways. Most frequently we manipu-lated the functional stimulus-processing time by varyingthe interval between stimulus and mask (stimulus-to-mask onset asynchrony; SOA). With practice, the minimalSOA required for threshold detection (e.g. achieving 80%correct) was substantially reduced [17].

The orientation specificity of this improvement indeedfollowed a consistent pattern: when stimulus detectionwas made difficult by a brief processing time (short SOA),small target/distractor orientation difference, increasedtarget position uncertainty, or a target presented fartherfrom fixation, improvement was slower and orientationspecific. By contrast, learning of easy cases was generaland transferred across orientations. This pattern wasfound both across groups training under easy or difficultconditions, respectively, and across easy and difficultconditions within the same individuals. Thus, practicingorientation detection with interleaved SOAs, observersshowed generalization for long (easy) SOAs and specificityfor short (difficult) SOAs. Moreover, when learning wasorientation specific, it was also position-specific, consist-ent with the linkage of orientation and position specificity

Figure 1.A hint for ‘understanding’ the image in Box 1. A few lines draw attention tothose aspects that are essential for perceiving a bearded figure. Now return to Box 1on the preceding page to see how this ‘clue’ has affected your perceptual system,perhaps forever!



EUREKA!

The reverse hierarchy theory of visualperceptual learningMerav Ahissar1 and Shaul Hochstein2

1Department of Psychology and Interdisciplinary Center for Neural Computation, Hebrew University, Jerusalem, 91905, Israel2Department of Neurobiology and Interdisciplinary Center for Neural Computation, Hebrew University, Jerusalem, 91905, Israel

Perceptual learning can be defined as practice-inducedimprovement in the ability to perform specific percep-tual tasks. We previously proposed the Reverse Hier-archy Theory as a unifying concept that links behavioralfindings of visual learning with physiological andanatomical data. Essentially, it asserts that learning isa top-down guided process, which begins at high-levelareas of the visual system, and when these do notsuffice, progresses backwards to the input levels, whichhave a better signal-to-noise ratio. This simple concepthas proved powerful in explaining a broad range offindings, including seemingly contradicting data. Wenow extend this concept to describe the dynamics ofskill acquisition and interpret recent behavioral andelectrophysiological findings.

Throughout life our sensory receptors are continuouslybombarded by stimuli. This activation not only inducesperception, it also modifies our representation mechan-isms, thereby affecting all subsequent perception. Recentevidence shows that a large degree of perceptual plasticityis retained in adulthood, with long-term manifestationsincluding adaptation, priming and perceptual learning.Currently, these different phenomena are defined by theirbehavioral characteristics and the manner in which theyare induced, rather than by their respective underlyingneural mechanisms.

This article focuses on perceptual learning, defined aspractice-induced improvement in the ability to performspecific perceptual tasks (see [1] for a classic review and[2] for recent overviews). We shall argue that whattypically limits naı̈ve performance is the accessibility oftask-relevant information rather than the absence of suchinformation within neuronal representations [3]. We shallpresent the reverse hierarchy theory (RHT) of perceptuallearning asserting (i) that perceptual improvement lar-gely stems from a gradual top-down-guided increase inusability of first high- then lower-level task-relevantinformation, and (ii) that this process is subserved by acascade of top-to-bottom level modifications that enhancetask-relevant, and prune irrelevant, information (see [4]for a computational model applying a similar concept).

The relations between plasticity processes duringsubstantial practice and those dominating the first fewexposures, are not well understood. Only the first are

typically referred to as perceptual learning, whereas thelatter are termed priming. Recent evidence suggests that,at least when governed by top-down control, singleexposures (priming) can induce strong and long-lastingeffects that clearly change our perception (see Box 1),suggesting that these might be the initial processes ofperceptual learning, as described below (TheEureka effect).

Box 1. The Eureka effect

The picture (Figure I) generally appears simply as a set of gray andblack regions, without further meaning; the object represented ishard to categorize without further cues. Following considerableinspection, (or a single look at the ‘clue’ in Figure 1 overleaf), thepuzzle is solved, and, without further practice, observers directlyperceive a bearded figure. This effect is rapid, strong and long-lasting, suggesting that significant top-down control determines ourconscious perception. Ahissar and Hochstein [21] found that a singlelong exposure to a ‘pop-out’ stimulus enabled learning of a verydifficult detection task, based on brief and strongly maskedpresentation of similar stimuli – a task that was almost never learnedwithout the Eureka enabling experience. Thus, it appears that similartop-down control or guidance mechanisms influence both percep-tual learning and conscious perception. An important differencebetween these effects is that in the experiment, following the singleeasy-case ‘Eureka’ exposure, hard-case perceptual learning wasenabled, but still required; no such training is needed for Figure Ihere.

Figure I.Corresponding author: Merav Ahissar ([email protected]).

www.sciencedirect.com 1364-6613/$ - see front matter Q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2004.08.011


Spatiaalinen avaruus

Escherin kuvat

Aivot eivät muodosta “yhtä kuvaa”, siksi osat tuntuvat mahdollisilta, vaikka tarkempi suhteiden vertailu tekee hahmosta mahdottoman.

Thatcher illuusio

Piirteet analysoidaan yksittäisinä suhteessa muistiainekseen

Vihannespuutarhuri

Parietaali-syndroomat

Optinen ataksia

Konstruktionaalinen apraksia

Hemispatiaalinen neglect

Balintin syndrooma

Ideomotorinen apraksia

64

Optinen ataksiaPotilaalla on vaikeuksia visuaalisesti havaittujen esineiden tavoittelussa

Voi lähteä oikeaan suuntaan, mutta menee sitten “kahmimiseksi”

potilas saattaa kyetä arvoimaan suuntiaei siis ainoastaan havaintohäiriö

65


Vaikeuksia rakentaa tai tuottaa kaksi- tai kolme-ulotteisia kuvioita

66RHD06

LHD08

MODMODEELL

RHD01

LHD03

MMOODDEE LL

RHD10

LHD01

MMOODDEE LL

RHD05

LHD10

MOMODEDE LL


Vasemman ja oikean aivopuoliskon vaurioilla eri oireet (local-global)

67

A of M's

S of J's

G of K's

D of Y's

LHD LHDRHD RHDA. B.


Todennäköisesti monisyinen häiriö

Useita eri tekijöitä

spatiaalis-havaintopohjaiset

motorisen suunnittelun

visuo-motorinen integraatio

sarjallinen suunnittelu (eksekutiivinen kontrolli)

68


neglect =huomiottajättäminen

Vaikeus havaita, reagoida tai suuntautua ärsykkeisiin, jotka esitetään kontralateraalisti vauriopuoleen nähden

Vasen neglect yleisempää kuin oikea

69


70Line cancellation task Line bisection task

Figure copy task


71

Koskee niin havaintoa kuin mielikuvia

Mikä luku on 12 ja 18 välissä keskellä? (Zorzi ym. 2003)

Fellini

Balintin syndrooma

Optinen ataksia

Simultagnosia = pystyy kiinnittämään huomiota ainoastaan yhteen ärsykkeeseen kerrallaan

molemminpuolinen okkipito-parietaalivaurio

72

Ideomotorinen apraksia

Kykenemättömyys suorittaa opittuja liikkeitä

Suurimmat vaikeudet suoritusten pantomiimissa (sahaaminen, puhelimeen vastaaminen), paranee imitaatiossa, paras suoritus oikeilla esineillä

Vaurio: vasen IPL tai SMA

73

Tutkimuksesta

Testaamisesta: Tunnistaminen

Visuaalinen identifikaatiopiirroskuvien tunnistaminen (esim. Bostonin nimeämistesti)oikeiden esineiden tunnistaminenkuvien ja niitä esittävien esineiden yhdistäminen

Kasvojen tunnistaminentutut kasvottuntemattomat kasvot

Yleisten symbolien tunnistaminenlogot, numerot, kirjaimet

75

Testaamisesta: Piirteiden erotteluSuuntaorientaatio

esim. viivojen suunnat

Kuvio-tausta -erotteluMuodontunnistusPäällekkäiset kuvatFragmentaariset kuvat

epätarkatosittaiset

Liikkeen ja liikkeen suunnan hahmottaminen

76

Testaamisesta: Suunnat ja tilat

Vasen -oikea erotteluitsestä ja toisesta

Käsitteiden hallintapäällä, sivulla, takana, edessä, jne.

Paikan muistaminensuhteessa tilaan: piilotustehtävätsuhteessa toisiin esineisiin (kuviosarjan muistaminen)suhteessa itseen

Kartan piirtäminenesim. oma huone, tutkimushuone, piha, luokka, Suomi

77

Testaamisesta: Konstruktiot

Kompleksien kuvien havainnointikuvasta kertominenreitin piirtäminen, kertominen

Etsi samanlainenvisuaalinen etsintämuistipelikortit

PiirtäminenReyn kuvio, ihminen, polkupyörä, kreikkalainen risti

Yksinkertaiset palapelit

78

Testaamisesta: Muisti

Työmuistikuviosarjojen muistaminenCorsi

Mallista piirretyn muistaminen välitönviivästetty

Toistuvat kuviotKuviomuistitestiMerkkikoe (merkkien muistaminen)

Piilotetun tai poistetun muistaminenKim -leikki

79

Testaamisesta: Suunnittelu

Palapelien tekeminen ja muu kokoaminen (lego-rakennelmat)KuutiotehtävätPäättely visuospatiaalisten suhteiden perusteella

Ravenin matriisit

PiirtäminenReyn kuvio

80

Spatiaalisuus ja motoriikka

Parietaalialueet ja frontaalialueet (motoriset alueet) erittäin tiukasti yhteydessä toisiinsa

Missä-prosessointi hyvin pitkälle Miten-prosessointia

Tavoittelu, liikkuminen, tasapaino, kaikki motoris-spatiaalisia prosesseja

81

Kiitos

Documents

Visuo- spatiaalinen - Jyväskylän yliopistousers.jyu.fi/~prasanen/Files/Visuospatiaaliset taidot.pdf · Paikkasolut Paikkasolut prosessoivat spatiaalista muistia siitä, missä jokin