Attention

Attention several different functions

For Filtering

Bottleneck to protect serial procssor

Early Filtering: Broadbent

Completely

Broadbents Filter Theory

Need to protect a serial processor overflow

We can only attend to one thing ("gate") at a time

Evidence

Broadbent: Dichotic Listening: Dont remember anything from unattended ear

but echoic store: If short messages: remember last words

but remember if voice m/f

Not Completely

Treisman

Attention for Attentuation (Turning Down a Signal)

Evidence: ERPs

Brain Stem: Normal Signal. Cortical: less magnitudeattending away from a stimulus actually reduces the intensity of the signal in the brain

What about unattn. Stim?

We do process meaning!

dont need much attn to be recog

Reason=Priming

Shadowing Evidence

Meaning Cues: Little red riding h continued on the other ear even when shadowed

Cocktail Party Effect

Meaning Cues Corteen and Wood: Response to unattended "Shocked words"

Neurological Evicence

Masking Evidence

Evett & Humphreys: Invisible Lion Primes Tiger

Attn can follow simple cues to shift

To overcome limited number of neurons

Lamme / DeLollo

receptive fields of neurons of higher visual areas seem to "shrink" around biased stiumuls

Late Filtering: Deutsch and Deutsch were wrong: Not Everything is processed fully

Attn for Binding

To combine features

Treisman

We need attention to combine features into objects

Cross modal Binding

ventriloquism effect: sound seems to come from lips in front although speakers on side

To enable for concious report

Coltheart

Consious generelly only when Combining Semantic (general) and Epsisodic (detail)

--> Attn joining two parallel proceses

Dorsal and Ventrals (What, Where) streams in visual processing

Patient DF: It has been suggested (Goodale and Milner)ventral system recognizes objects. Object Related! The dorsal system more in driving some action in relation to an object. Viewer Related

Evidence: When pushed in Sperling they report letters but no colours or loc.

Attention for action:

Allports Berries:

We cannot pick more than 2 berries

Attn. fits possible actions

Several Modules are constantly processing input

Many things are processed in parallel but not fully

Meaning rather than specific Epsisode

An attentional filter is constantly surveying input channels for important material

Attenion can be biased

Salient Stuff

Current Goals / needs

Only if attended to, matrial can be selected for concious report

So many things might be processed automatically to an even categoric level (Lamme)

Stages of Processing

Attention needed

Input

Transduction

Photons to Chemicals to Electircal by Receptors

Perception

Storage

Iconic Memory (Phenomenal Awareness)

Recognition

Contains highly processed stuff

However only attened stuff is fully procesed (meaning + particular look of instance)

Rest Primes but decays

Short Term Memory

Access Awareness

Long Term Memory

Declarative: Concious

Semantic

Episodic

Procedural: Not-Concious

Output

Thinking

Reasoning

Language

Non-concious Material: Stuff that we have no receptors for

Genetic / Ecperience Depenent

Attention as LTM /learned system bias

Lamme: Stimulus driven: size, contras, movement etc

Situation Dependent

Bottom up (Stimulus Driven)

Orientation reaction

Top Down (Goal Driven)

Visual Search

Overt

By Eye Movements

Covert

By changig neural acitvy (shrinking / enancing receptive field)

Contious / vs. Partial

Object / Location based / Semtincs based

Can follow story ear to ear

Firure / Ground selection

Attention tying together different modalities

Ventriloquism Effect

Location dependent

Zoom in / out

level of detail gets worse

Sinlge

Devided

Unfocused

Visual

Auditory

We can proces the stiumulus but not be concious of it: no Access Concious

Koch: Attention as mechanism that could be modeled: No consciousess needed

Coltheart: A Masked Stimulus is fully processed but: Uniting episodic & Semantic required for consciousness

Naish: "Attention gives rise to concious processing"

Attn needed for Perception? No!

Attn to glue features into objects

Treisman

Binding Features (Inital Perception as "Soup of features")

"Preattentive" Processes vs. Focal Attn.

Evidence

Patient with pariatal damage wrongly binds features

POP-Out Effect

Paralell search when target and distractors discriminated by single feature

Serial Search (Attn)

Conjunction Search (Featrue binding is neccessary)

Target

Doesn't pop out

Distractors

Distractors need Attn

No. of Distrators plays role

Parallel Search (no Attn)

Target

Target pops out

If only single feature no Attn required

Distractors

No. of distractors play no role

More complex stimulus pop out

Today: "Efficient Search"

Duncan + Humphreys: L+T (L should still pop-out)

Flanker-Effect should not be possible: Words need binding - However Attn only only on middle word

Negative effects of short term exporsure

Sperling (50ms)

3-4 letters from 12 when cued by after signal

->Iconic memory

Lamme

Phenomenal Conciousness

Attention by simple cues: colour and position (tone cue)

If attended: Enter WM

Only Simple cues: not "attend to vowels "

Attn for concious report? Yes!

DiLollos& Gisbrechts

Two-Stage Model

1. Parallel processing of features (Identity (Semantic), Size. Location, Shape, Colour(Episodic)

2. Serial (Attention) Processes to join Sem + Epi for concious report

While Stage 2 is engaged , made difficult later information cannot be processed, so has to remain at Stage 1

Evidence

RSVP: If stim. replaced /decayed= nothing to attend to possible

RSVP: Epsiodic info morea easily overwritten

If the second customer chats to behind forgets what he came for (info overwritten)

Ennns & DiLollo:

Re-Entrant Model of Processing

Bottom-Up Parrallel: Generates highlevel Hyptothesis

Top-Down: Hypothesis confirmation

Evidence:

Neurological

Recurrent Processes are stopped

Observational

Attentional Blink

No awareness of secnd T if presented shortly after first

Masking effects

Enns& DiLollo: Masking: With four dots around -isn't expected

Other Evidence

Flanker Effect (Shaffer and LaBerge)

It takes longer to decide ‘dog’ is an animal when surrounded by words of another category

Non attended words are processed - Att. not needed to join features

Broadbent saves Treisman

Cat is not percieved - its primed, or it takes only minimal Att because of priming

Neglect

Attend only onle side of the visual field

Object Based

Half Clock drawn

Location Based

Half Plate is eaten

Balint Syndrome

Able to see only one thing at a time

Unable to shift attention

Damage to pariatal lobe

Depression

Attentional Bias: Emotional Stroop: Take longer to name the colour of anxiety related words

Lamme:

Attention as guiding feed forward sweep of information

Weights of stored knowledge of the system (genetics, salience, needs)

"Last Traces"

Objects in receptive fields compete for processing limited amount of neurons

Koch: conciousness can be seen as the outcome of competitions / alliances / cooperations between biased neurons