Pattern Recognition,
Attention, and Consciousness

I.                   Pattern Recognition: Vision

a.       Definition
– identification of a complex arrangement of sensory stimuli

b.      Template-Matching
Model

                                                              i.      Templates: stored (specific) models of all recognizable patterns

                                                            ii.      Compare item of interest with templates of items stored
in memory

                                                          iii.      To identify an object, every aspect of the object must
match the template perfectly

                                                          iv.      Problem: there is enormous variability in patterns that we can recognize

1.      Requires
an infinite number of templates for any given object

2.      Object
identification would take a long time

c.       Prototype
Model

                                                              i.      Stores a prototype of an object or pattern

1.      Prototype:
abstract, idealized patterns

                                                            ii.      Compare a target object with our existing prototype and
identify an object based on closest match

1.      Tolerate
minor variations between the object and prototype

2.      Where
do the prototypes come from?

d.      Distinctive
Feature Model

                                                              i.      Feature: a very simple pattern: a fragment or component
that can appear in combination with other features

                                                            ii.      Recognize the whole pattern by breaking objects apart

                                                          iii.      Hubel and Wiesels’ research

1.      In
the cortex of a cat, physiological evidence for special receptors that respond
to only certain features

2.      Recognize
an object as a result of activation of a number of specific cells

                                                          iv.      To indentify an object, it is not just the features
that influence recognition, but also the features background

1.      Nontargets

2.      Duncan
& Humphreys showed that features of the nontargets influence pattern
recognition

3.      We
can shut-off some features to increase recognition speed

4.      It
is more difficult to locate a target when it shares features with nontargets

                                                            v.      Top-down processing

1.      Conceptually
driven processing: higher cognitive processes assist in pattern recogintion,
cognitive processes alter visual recognition

a.       So,
how might top-down processing influence pattern
recognition within the model?

                                                                                                                                      i.      The more likely the letter, the more activation and the
greater weight assigned to that letter

                                                                                                                                    ii.      Higher level knowledge and expectations assist your
processing of perceptual information

                                                                                                                                  iii.      Word superiority effect:

                                                                                                                                  iv.      Faster at identifying a single letter within the
context of word than when the letter is presented alone

1.      Each
letter sends an excitatory sign to other letters of a familiar word

                                                                                                                                    v.      How does the model explain face recognition?

1.      When
all the familiar features of a face are prsent, you are faster at identifying
even a single feautre because of excitatory signals sent to that feature from
other features.

                                                          vi.      Bottom-up processing

1.      Data
driven processing: using only features and clues in the target stimulus to
identify the object

e.       Computational
Approach

                                                              i.      We perform computations on incoming stimuli; the
products of the computation ultimately permit pattern recognition

                                                            ii.      Recognition by components (RBC) model

1.      Irving
Biederman

2.      Identify
3 dimensional objects by recognizing the component geons and arrangement of
geons (geometric icons)

f.       Connectionist
Model (neural networks, parallel distributing processing)

                                                              i.      Compute an object based on both bottom-up and top-down
processing

                                                            ii.      Three levels

1.      Input
level – cells receive input from the environment via feature detectors

2.      Hidden
units – the units receive either positive activation that leads to “excitation”
or negative activation that leads to “inhibition”

a.       “internal”
processing

3.      Output
units

a.       Connections
from all parts are computed simultaneously

b.      Response
in whatever form is made as a result of registering multiple computations

                                                          iii.      Word superiority effect

                                                          iv.      Face recognition

g.      Gestalt
Psychology

II.                Pattern Recognition: Audition

a.       For
vision, you receive information across the visual field or across space

b.      For
audition, you receive information about sounds in the environment across time

c.       Template
and prototype explanations for pattern recognition in audition have been
rejected

d.      Distictive
feature approaches do not account for top-down influences

e.       Conceptually
driven recognition – top-down processing

                                                              i.      Warren and Warren’s research

1.      It
was found the *eel was on the axle

2.      It
was found the *eel was on the shoe

3.      When
Ss were asked to repeat exactly what they heard, the reported the sentence
replacing the cough with the phoneme to complete the word

4.      Most
Ss reported no awareness of the cough at all.

f.       Selective
attention in auditory perception

                                                              i.      Selective attention – when one is confronted with
multiple tasks and must attend to one task while disregarding other tasks

1.      Cocktail
party effect

                                                            ii.      Filter theories – theories that suggest we select out
an important message and filter out unimportant messages

1.      dichotic
listening task – while wearing earphones, Ss are presented with two different
messages one to each ear

                                                          iii.      Cherry’s research – 1953

1.      Shadowing
requires Ss to repeat the message heard in one ear aloud as it is being said
while ignoring the other ear’s message

2.      Ss
could do this task fairly well, but tended to shadow the ear in a monotone.

3.      Altered
the message to the unattended ear after the Ss had been shadowing for a while
and found the following:

a.       Ss
could detect a difference in the background message when it changed from speech
to tone

b.      Noticed
when the voice in the unattended ear changed from male to female and vice versa

c.       Did
not notice when the voice in the unattended ear switched languages

d.      Could
not indentify any words or phrases

                                                                                                                                      i.      Even when a single word was repeated 35 times

e.       Suggest
we do monitor some aspects of the unattended channel, like pitch

g.      Selection
models

                                                              i.      Broadbent’s early selection model –

1.      Selection
occurs in the initial stage of information processing

2.      Selection
is based upon physical features like tone, loudness, location, pitch, etc.

3.      Problems:

a.       Treisman
(1960,1964)

                                                                                                                                      i.      In one study, she used the same speaker for both messages

1.      Cannot
filter the message based upon physical characteristics of the speaker

2.      Although
more difficult, Ss could still shadow the one message

                                                                                                                                    ii.      Treisman concluded that we select or filter out info
based upon the content of the message

                                                            ii.      Stage 2 selection model

1.      At
this stage, selection is based on grammatical and semantic features of the
stimulus

2.      Treisman’s
evidence

a.       After
Ss shadowed a few words of the message in one ear (with the same voice), the
message was switched to the unattended ear

                                                                                                                                      i.      Ss continued to shadow the same message, but now in the
ear they were told to ignore

                                                                                                                                    ii.      Ss were not aware the message had switched ears

b.      We
are not selecting out as early as Broabend suggested

c.       Treisman
believed all messages receive some low level analysis

d.      Selection
for a single channel occurs at the semantic level

                                                          iii.      Stage 3: late selection model

1.      Deutsch
and Deutsch’s research

2.      All
stimuli are completely analyzed and that selection occurs at the level of
responding

III.             Attention: Mechanisms, Processes, and Mental Resources

a.       Automatic
and conscious processing

                                                              i.      Automatic processing (divided attention by Treisman)

1.      Stroop
effectm – green written in blue

a.       Priming:
when there is automatic access to info in memory

b.      For
the Stroop Effect, letters provide automatic access to the meaning of the word

c.       Fits
with the connectionist/PDP models

2.      Not
part of concious awareness

a.       You
are aware of the products of processing

b.      Not
aware of how this information
was made available

3.      Consumes
little or no process

a.       Does
not interfere with other processes

                                                            ii.      Controlled processing (focused attention by Treisman)

1.      Occurs
only with intention

2.      Open
to awareness

3.      Requires
attentional resources

a.       Can’t
do more than one task

b.      Serial
processing of info

                                                          iii.      Practice and memory

1.      With
practice, a complex and attention demanding task becomes easy, less resource
demanding

2.      Practice
helps you store info in memory about how to perform the task more efficiently

a.       Shiffirin
& Schneider

3.      Synthesis
of perception and attention

a.       Limited
amount of attentional resources

b.      We
can focus our attention on a single demanding task or divide our attention
among 2 less demanding tasks

c.       Automatic/controlled
processing occurs on a continuum

b.      Synthesis
of attention and memory

c.       Illusory
conjunctions in attention

                                                              i.      Illusory conjuctions: inappropriate combination of
features

                                                            ii.      Treisman and Gelade’s research

1.      Spotlight
attention/focused attention

2.      Asked
Ss to search a visual display for either a simple feature or a conjunction

a.       Simple
features were found rapidly, regardless of the display size (number of
nontargets)

                                                                                                                                      i.      Spotlight permitted search of the entire display at the
same time

b.      When
engaging in a conjuction search, display size was a factor

                                                                                                                                      i.      Took more time as the number of nontargets increased

                                                                                                                                    ii.      The spotlight has to look at one letter at a time

                                                          iii.      Physio support

1.      Feature
inhibitory mechanism: when features of nontargets are very different from the
target, we simultaneously inhibit irrelevant distractors

d.      Biology
of attention

                                                              i.      Relevant pathways

                                                            ii.      Methodology