Cognitive Neuropsychology Methods

1
Cognitive Neuropsychology Methods

• Aims and Objectives
• By the end of this lecture you will have learned
• The key methodological approaches used in
  cognitive neuropsychology
• The importance of double dissociations in
  cognitive neuropsychology
• The main arguments on both sides of the
  single-case vs group study debate
• Required Reading
• Caramazza, A (1986) On drawing inferences about
  the structure of normal cognitive systems from
  the analysis of patterns of impaired performance
  Brain and Cognition, 5, 41-66.
• Bub, DN (2000). Methodological issues confronting
  PET and fMRI studies of cognitive function.
  Cognitive Neuropsychology, 17, 467-484.

2
Cognitive Neuropsychology Methods

• Associations
• Dissociations
• Double Dissociations
• Single case vs group studies
• Functional neuroimaging.
• Neural network modelling
• Animal studies

3
Associations

• An association implies a link or connection
  between two phenomena.
• Between two cognitive deficits (e.g.
  comprehending written and spoken words)
• Between a cognitive deficit and a lesion site
  (e.g. left hemineglect and right parietal lobe
  lesions)
• Problems - cant determine causality, (cf
  correlation) nearly always exceptions found.
• Association may occur for biological rather than
  cognitive reasons.

4
Dissociations

• Patient A Performance on task X impaired, but
  performance on task Y intact
• Performance on tasks X and Y dissociates
• E.g. task X word recognition, task Y face
  recognition
• Implication is that face recognition and word
  recognition are handled by different sets of
  cognitive processes, and only the word
  recognition system is damaged in patient A.

5
Faces
Recognition Module
OUTPUT
Words
6
OUTPUT
Faces
OUTPUT
Words
7
Dissociations

• Shallice (1988) described 3 types of dissociations

• Classical dissociations presumed to be the most
  powerful

8
When is a dissociation a dissociation?

• Crawford et al (2003) - What is a dissocation?
• E.g. what do we mean by impaired, or deficit ?
• Typical statistical procedures not necessarily
  appropriate for single case studies.
• Standard procedure is to use Z-scores - but not
  good with small control groups
• Crawford et al advocate modified t-test to
  determine whether patients performance is worse
  than controls
• BUT what do we mean by within normal limits?
• To satisfy this we must prove the null hypothesis
  (when we can only fail to reject it)
• SO redefine normal as fail to meet impaired
  criterion.
• Finally, performance on task X and task Y must be
  significantly different (by modified Z-test)

9
Dissociations

• Interpretation of dissociations is not always
  straightforward.
• It could be argued that tasks X and Y involve one
  process (e.g. recognition of "something") but
  that word recognition is a very hard task and
  face recognition is a much simpler task.
• Maybe brain damage affects difficult tasks first?
• Task difficulty effect / resource artefact

10
Faces (20)
Recognition Module (100)
OUTPUT
Words (60)
11
Faces (20)
Recognition Module (50)
OUTPUT
Words (60)
12
Double Dissociations

• But
• Patient B Performance on task X intact, but
  performance on task Y impaired
• E.g. Facial recognition impaired but word
  recognition intact
• The performance of patients A B provide a
• DOUBLE DISSOCIATION
• Strong evidence that there are cognitive
  processes involved in Task X that are not
  involved in Task Y and vice versa
• Patients don't have to be perfectly intact on
  either task - they just need to be consistently
  better at one task than the other

13
Double Dissociations
Normal range
Performance
CLASSICAL
STRONG
Other types of dissociation, such as uncrossed,
also occur
14
How important are DDs?(What can be inferred from
DDs? Cortex, 39(1) 2003)

• DDs traditionally assumed to be gold standard
  in CN research BUT - not all CNs agree -
• Caramazza argues that associations, dissociations
  and DDs are all equally valid forms of inference
  (if the cognitive model is well developed)
• A DD between two tasks does not necessarily imply
  a DD between cognitive processes (Shallice, 1988)
• E.g. lesioned neural network models with no
  obvious modular structure can produce data that
  looks like a DD (Plaut, 1995).
• The utility of DDs is predicated on modularity
  being true - Van Orden et al, 2001 Endless
  fractionation
• Most CNs agree that converging evidence is
  desirable

15
Single case vs Group studiesThe concept of
syndromes

• Traditional neuropsychology often based on
  syndromes - a collection of symptoms which
  often co-occur in individuals. Early syndromes
  were anatomically based (e.g. Brocas Aphasia)
• Gerstmanns Syndrome Acalculia, left-right
  disorientation, pure agraphia, finger agnosia
• Can the study of GS provide information about
  the
• functional architecture of cognitive processes?
• Association of deficits on these tasks implies
  they share an underlying process
• Requires a model with a component common to all
  symptoms

16
Single case vs Group studiesThe concept of
syndromes

• The role of such a processing module is not
  obvious
• It is more likely that these symptoms depend on
  a number of functionally distinct processes which
  are anatomically related. (Danger of
  over-interpreting associations)
• This is one reason why some cognitive
  neuropsychologists favour single-case studies
  over group (syndrome) studies
• Research based on classical syndrome types
  should not be carried out if the goal of the
  research is to address issues concerning the
  structure of cognitive processes Caramazza
  (1984)
• In other words, classical syndromes based on
  anatomical considerations have no role in
  cognitive neuropsychology

17
Single case Vs Group studies

• Caramazzas arguments
• ONLY the single-case approach can provide
  information relevant to our understanding of
  cognitive architecture
• WHY?
• Group studies rest on assumption that cognitive
  processes are homogenous (patients grouped to
  minimise sampling error - noise)
• BUT Brain damage may disrupt cognitive processes
  in a variety of different ways
• Therefore performance differences within a group
  of brain damaged subjects CANNOT be dismissed as
  noise.
• Therefore averaging over a group of patients is
  inappropriate

18
Single case Vs Group studies

• Caramazzas arguments
• Negative consequences of averaging
• 6 patients, 6 controls, both groups do 2
  neuropsychological tests

Conclusion - Patients impaired on tasks X and Y
19
Single case Vs Group studies

• Caramazzas arguments
• Negative consequences of averaging
• Group differences may not reflect performance of
  any patient.

No individual patient is impaired on both tasks 2
different groups?
20
Single case Vs Group studies

• One response is to study Functional Syndromes -
  based on IP models of normal function
• E.g. specify criteria on basis of cognitive model
  which will identify a group of patients who are
  homogenous with respect to the proposed cognitive
  impairment
• E.g. deep dyslexia, surface dyslexia,
  phonological dyslexia
• BUT -
• Patients may be homegenous with respect to
  task(s) used to select them, but not with respect
  to experimental task.
• Selection criteria often poorly specified /
  theoretically weak

21
Single case Vs Group studies

• Objections to Caramazzas position
• The same logic may be applied to the study of
  normal behaviour, resulting in the rejection of
  group studies throughout psychology. (Shallice,
  1988)
• Single cases may simply be the most extreme
  examples of a larger, ignored group. (Robertson
  et al , 1993).
• Single cases make establishing brain-behaviour
  relationships difficult. (Robertson et al ,
  1993).

22
Single case vs Group studies

• Other arguments against a single case only
  position
• single case studies cannot address theories to do
  with relationships between two variables (e.g.
  brain size and intelligence) since correlational
  designs need many subjects
• Single case studies do not permit pure
  replication
• patients can sometimes be atypical from the
  outset (Caramazza's "martian within us" problem),
  e.g. split-brain patients whose brains have
  developed non-conventionally.

23
Single case vs Group studies

• The debate is very detailed and complicated, (and
  ongoing)
• E.g. Caramazza Badecker (1991) Clinical
  syndromes are not gods gift to cognitive
  neuropsychology - a reply to a rebuttal to an
  answer to a response to the case against
  syndrome-based research. Brain Cognition, 16
  211-227
• The debate addresses many of the assumptions
  outlined in Lecture 1.
• It has involved philosphers as well as cognitive
  neuropsychologists
• cognitive neuropsychology practice not only must
  steer clear of the Scylla of sole reliance on a
  standard reductionist approach that relies solely
  on group studies, but also would do better to
  avoid the Charybdis of ultra-cognitive
  neuropsychology Shallice, 1988.

24
Functional Neuroimaging

• Many different techniques, eg
• SPECT
• PET
• fMRI
• TMS / rTMS
• EEG
• MEG
• Currently fMRI, TMS and MEG are the most
  popular techniques
• Techniques are increasingly combined

25
Functional Neuroimaging
26
Functional Neuroimaging

• Utility of functional neuroimaging for cognitive
  neuropsychology
• Field is largely split
• PROs
• Can potentially localise function in healthy
  controls
• Has revealed activity in brain areas previously
  thought to be uninvolved (e.g. cerebellum)
• CONs
• Interpretation of imaging data not
  straightforward
• No standard vocabularly for describing results
• Replication of results often poor
• Does nothing for theory development
  (Parkin,2001).

27
Computational Modelling

• Generally uses connectionist (PDP) architectures
  to model aspects of cognition.
• Models are built, and then lesioned in various
  ways (e.g. units knocked out, weights changed)
• Model a success if resulting output resembles
  patterns observed in brain damaged humans.
• Advantages
• Non-invasive
• Forces researchers to specify cognitive theory
  adequately.
• Disadvantages
• Biological plausability unclear

28
Animal studies

• Many (non-language) cognitive functions are also
  studied in animals
• E.g. Memory, Attention, Executive functions.
• Advantages
• Discrete, replicable lesions (permanent or
  temporary)
• Age, environmental effects controlled
• Single cell recording
• Neuropharmacology of cognition
• Disadvantages
• Not very nice for animals
• Unclear how far data can be generalised to humans

29
Summary

• Cognitive Neuropsychologists make inferences
  about the architecture of cognition on the basis
  of assocations, dissocations and double
  dissocations.
• These inferences are only valid if certain
  assumptions are true
• Single case studies are a very important source
  of information for CNs (but not the only source)
• The pros and cons of single case studies are
  still debated
• Modern functional neuroimaging techniques are
  increasingly used -
• But the same assumptions are required in order to
  interpret neuoimaging data, and the techniques
  have other additional drawbacks.