Neuro vs. Cognitive Psychology:

1
Neuro vs. Cognitive Psychology
A case study
2
Outline

• What is activation?- The view from fMRI
• The logic of subtraction
• Imaging orthographic similarity

3
Signs of activation

• Cellular activity in the brain is accompanied by
• Increased blood flow (and so temperature) in the
  activated area
• Increased oxygen uptake in the activated area
• Increased glucose use (during oxidative
  metabolism)
• IF we can detect changes in blood flow or oxygen
  uptake or glucose metabolism or temperature, then
  we can deduce where cellular activity differences
  occur during any given task

4
Functional magnetic resonance imaging (fMRI)

• Measure blood oxygen level dependent (BOLD)
  signal
• - increased local CBF during activity leads to
  excessive oxygenated hemoglobin (oxyhemoglobin)
  in that region (Anyone know why?)
• - Oxygenated and deoxygenated hemoglobin have
  different magnetic properties, the latter being
  magnetically charged
• - We can detect two different relaxation times
  T1 (spin lattice relaxation time) and T2
  (spin-spin relaxation time)- it is the latter
  that is used for functional imaging
• T2 is induced by local magnetic field
  homogeneity in the slice under current study
• fMRI resolution is about 1 X 1 x 3-4 mm.
  temporal resolution is several seconds for whole
  brain

5
Subtraction logic

• Due to Donders, 1868
• Lets say you are interested in A
• Devise a task AB, which incorporates A
• Ask subjects to do AB and B alone
• Then (AB) - B Time to do A
• i.e. Color discrimination of lights - RT to
  lights Color discrimination time

6
Subtraction logic

• Subtraction logic makes many assumptions, some of
  which are debatable
• The subtraction method necessarily or implicitly
  assumes
• that cognitive processing is serial
• that cognitive processing is hierarchically
  organized
• that cognitive processing unfolds in an
  exclusively forward fashion
• that structures participate in an all or nothing
  fashion in a cognitive process

7
Subtraction logic in fMRI

• The subtraction method necessarily or implicitly
  assumes
• that peak CBF or glucose uptake or oxygen use
  corresponds to one single cognitive component of
  the task
• that the same cognitive component of a task is
  always performed by the same brain region (and
  thus, implicitly, that the brain is not
  redundantly organized), even if that component is
  shared between different tasks
• that subjects perform all and only the requested
  task (or that other tasks are are associated with
  random or perfectly consistent activity)

8
Subtraction logic

• None of the assumptions seems terribly likely,
  and several fly in the face of current theory
  about brain organization.
• What can we do to overcome doubt? Gather
  converging evidence.

9
Subtraction logic

• Subtraction logic is almost always used in
  imaging experiments
• Recently, some have started using
  auto-correlation instead, but this is not
  wide-spread
• The nature and purity of the subtraction is
  vital to interpretation of the imaging results
• For this reason, imaging results and experimental
  design are intimately and necessarily yoked

10
Language studies

• Language access is very fast very complex, with
  multiple micro-functional constraints
• Experimental psycholinguistics has identified an
  over-whelming number of variables (several dozen)
  with demonstrable behavioral impact on lexical
  access multiple functional constraints in
  play
• There was a dissociation between early language
  imaging and psycholinguistic understanding, with
  stimuli in imaging studies failing to meet the
  rigourous control demands of psycholinguistic
  understanding

11
Micro-functional dissection

• Since even the simplest lexical access task is a
  multi-dimensional conglomeration of
  functionality, the key is to use very simple
  tasks, with very highly-controlled stimuli
• In this way we try to trap an automatic
  function of interest, well below conscious
  awareness
• And we pray that it is fine-grained enough to be
  informative!

12
ON

• J.R. Binder, K.A. McKiernan, M.E. Parsons,
  C.F. Westbury, E.T. Possing, J.N. Kaufman, L.
  Buchanan (in press) Neural Systems Underlying
  Lexical Access During Word Recognition, Journal
  of Cognitive Neuroscience.

13
ON

• Colthearts Orthographic N ON The number of
  words that are one-letter different from the
  target word
• -i.e. DOG ---gt HOG, DOE, DOT, DIG etc.
•  Many experiments manipulating ON have found a
  frequency-modulated neighborhood size effect.
• Uncommon words with large ON are recognized as
  words more rapidly than low-frequency words with
  small neighborhoods
•  This effect disappears with common words
• This is among the bigger effects, with freq
  and ON together accounting for gt 30 of variance
  in behavioral measures

14
(No Transcript)
15
Almost all words are uncommon.
16
The trap

• Task is lexical decision decide whether or not a
  presented string is a word
• 50 high/low ON concrete nouns nonwords (100
  each in all) matched one-by-one on frequency,
  length, bigram frequency, and phonological
  neighborhood size, and (between wordness) on ON
• the NWs are highly word-like, the two real
  word sets are very similar to each other except
  for the manipulated ON

17
Hypotheses

• (i) Words should produce stronger activation than
  word-like nonwords in many of the brain regions
  previously identified in studies comparing
  semantic to non-semantic tasks, and
• (ii) A subset of these regions should show
  stronger responses to items with many lexical
  neighbors, indicating activation of pre-semantic
  word codes.

18
Behavioral data RTs
Psych Lab
Scanner
19
Behavioral data Errors
Psych Lab
Scanner
20
fMRI parameters

• GE Signa 1.5 Tesla scanner
• T1-weighted anatomical reference images 124
  contiguous sagittal slices (.9375 x .9375 x 1.2
  mm)
• Functional imaging 19 contiguous (7 - 7.5 mm)
  sagittal slice locations covering the entire
  brain x 3.75 x 3.75 mm
• 136 whole-brain image volumes collected from each
  subject at 2-sec intervals
• Each image was yoked to a behavioral decision
  (event-activated fMRI), allowing separate imaging
  of high/low ON x W/NW

21
Words vs NWs
22
Words - NWs
23
Words - NWs
i.) Almost exclusively LH
24
Words - NWs
ii.) Dorsal inferior medial prefrontal activity
25
Semantic decision - phonological decision
26
Semantic decision - phonological decision
27
Words - NWs
iii.) Angular gyrus activity
28
Semantic decision - phonological decision
29
Transcortical sensory aphasia

• Damage to the long route between Brocas
  Wernickes area
• Main feature is a deficit in accessing (thinking
  about or remembering) the meanings of words
• - Comprehension is therefore severely impaired
• - The patient can neither read nor write and has
  major difficulty in word finding

X
Lichtheim, 1885
30
Words - NWs
iv.) Extensive midline activity
31
Words - NWs
iv.) Extensive midline activity
32
High versus low ON
33
Small ON hard - Large ON easy
34
i.) Small ON activation gt Large ON activation

• We had (perhaps foolishly) hypothesized the
  opposite
• Although small ON is harder by evidence of RT
  and error rates, high ON seems to coordinate a
  wider variety of information
• However Greater constraints Easier computation
  
• Think of 20 questions after 19 questions have
  been asked

35
ii.) Bilateral midline activity

• The midline is not normally associated with
  lexical processing

36
ii.) Bilateral midline activity

• The midline is not normally associated with
  lexical processing
• But we saw some in the W - NW contrasts

37
ii.) Bilateral midline activity

• The midline is not normally associated with
  lexical processing
• But we saw some in the W - NW contrasts
• And it was mirrored in the semantic tasks

38
iii.) Words gtgt NWs

• There is almost no activity for the high - low ON
  condition for NWs

39
iii.) Words gtgt NWs

• There is almost no activity for the high - low ON
  condition for NWs
• What differentiates words from NWs?
• Semantics!
• By evidence of activation, ON manipulations are
  sensitive to semantics

40
ON vs Semantics - phonology
41
ON vs Semantics
42
Semantics as a final push

• Small ON words seem to require more extensive
  semantic processing
• Why? To compensate for the fact that these items
  are less orthographically word-like.
• High ON biases the subject toward a positive
  response (increases the tendency towards yes)
• Relatively little semantic activation is then
  needed to complete the response selection task
  hence low ON - high ON looks like semantics -
  phonology
• This explains why low ON gt high ON, and why we
  dont see the effects for NWs, which take the
  same semantic processing in both ON conditions

43
Why are high ON NWs slow and error-prone?

• Presumably high ON NWs are rejected more slowly
  and more likely to be accepted because of their
  greater resemblance to words harder to reject
• However, the semantics interpretation fails no
  NWs have any semantics
• And we cannot explain why there are (almost) no
  imagable effects of this very reliable behavioral
  difference, save some puzzling midline activity

44
Cognitive Neuro psychology

• We probably would not (did not) have a view that
  ascribed semantic effects to ON sensitivity
  without imaging Experimentation fails
• However, the only evidence we have (right now) of
  ON effects in NWs are robust experimental
  effects Imaging fails
• It is a good thing that cognitive neuropsychology
  embraces both behavior and the brain

45
fin.