Methodology Of Cognitive Neuropsychology Research Paper

One still controversial example is that of the so-called ‘category-specific disorders’ in which the patient loses the ability to identify one particular class of item. Thus certain patients with herpes simplex encephalitis can virtually entirely lose the ability to identify animals, plants, and foods, but show a far better preservation of their knowledge of human artifacts (Warrington and Shallice 1984). This is a loss of knowledge and not of early perceptual processes or the lower levels of language. By contrast other patients can be considerably better at identifying the ‘animate categories’ than human artifacts (Warrington and McCarthy 1983). This double dissociation was originally interpreted in terms of the semantic representations of sensory quality knowledge being stored separately from knowledge about function, the latter being the more critical for identifying artifacts. Such an approach has been elaborated into quite complex models (e.g., Farah and McClelland 1991). Alternative ways of modeling the double dissociation have, how-ever, been produced (e.g., Gonnerman et al. 1997, but see Garrard et al. 1998). Moreover the original results have led to extensive functional-imaging investigations of the regions activated when different types of knowledge are concerned (e.g., Moore and Price 1999).

Very different methodologies have been used when patients are studied experimentally for what they can tell about the organization of cognitive functions. In the early twentieth century, when the nineteenth-century method of rather casual clinical descriptions was rejected for a more scientific approach, patients began to be studied as groups. Initially this took the form of a series of case histories (e.g., Head 1926), but in the 1950s and 1960s it became normal to average behavior over patients with particular characteristics. Groups could be defined by their gross lesion site (e.g., Milner 1958), or in terms of their global syndrome (e.g., in language fluent vs. nonfluent aphasia deriving from the work of Goodglass et al. 1964).

This approach has remained the standard one for those approaching neuropsychology with either a physiological psychology or clinical background. However in the 1960s and 1970s researchers more influenced by cognitive psychology began to use single-case studies increasingly because they were held to capture more satisfactorily from a theoretical perspective the essence of selective disorders corresponding to impairments of single cognitive processes. Indeed, in the 1980s it became popular amongst this group of researchers to argue that only single-case studies could be properly said to produce behavioral findings from neurological patients from which one could draw theoretical conclusions about the organization of the normal cognitive system (Ellis 1987). The position was best articulated by Caramazza (1986), whose argument was essentially that the behavior of the average of a group of subjects was not necessarily equivalent to that of any possible patient in the group. The argument will be discussed in more detail below.

One may now distinguish a variety of different types of experimental methods in regular use in cognitive neuropsychology. However I will begin with the simplest and arguably the most powerful method.

1. The Single-Case Study

In this type of study one selects for study a patient whose disorder appears to be in conflict with the implications of an existing theory, or to suggest a novel theoretical possibility. An example of the first type of situation is the study of patient KF (Shallice and Warrington 1970), where it was shown that the patient performed quite normally on a variety of tests of auditory-verbal long-term memory but was grossly impaired on measures of what would now be called the ‘phonological buffer’ (Baddeley 1986), such as the recency component of the (auditory-verbal) free recall serial position curve. This was in conflict with memory theory of the time, which held that the laying down of information in the long-term (secondary) memory system required that it be retained initially in a short-term (primary) memory system. The study showed that measures of auditory-verbal long-term memory could be normal even if measures of auditory-verbal short-term memory were grossly impaired, and was held to be in conflict with a serial organization of the short-and long-term verbal memory systems.

An example of the second type is the analysis of the writing disorder of patient LB (Caramazza et al. 1987). In this study it was shown that semantic variables did not influence LB’s rate of spelling errors, and also that lexical variables had little effect, but that performance was strongly dependent on word length. Moreover the errors could nearly always be interpreted in terms of single or double operations on single letters, such as substitutions, deletions, insertions, and transpositions. The pattern was held to support the existence of a graphemic buffer which held abstract letter-level in-formation while words were being written.

Studies of these sorts should involve either two or three stages following the stage of patient selection. First, one needs to carry out standard clinical (baseline) tests. The single-case study methodology differs from that in most other sciences in that replication of any given result is not generally possible for other re-searchers in the field. If a highly counterintuitive result is obtained with apparently major theoretical consequences, such as the initial observations of category-specific semantic memory dissociations (e.g., Nielson 1946) or the observations of object-based neglect described in patient NG (Caramazza and Hillis 1990), then it is not open to other workers in the field to attempt to replicate it. They generally do not have access to the patient. Therefore it is necessary that as rich a description of the patient’s general cognitive characteristics are given so as to provide the raw material for other researchers in the field to produce alternative explanations of any novel experimental findings. This is best provided by giving the patient’s performance on standard clinical tests in the particular cognitive domain.

Second, the investigation will involve tests used to examine particular theoretical possibilities for the patient’s difficulties. These can be obtained from the experimental psychological or neuropsychological literature or be specifically constructed. It is normally necessary that control data from normal subjects also be provided. Again because of the general impossibility of replicating investigations in findings in single-case studies, all theoretical inferences drawn from the patient’s performance should be based on the results of at least two experiments, or on a patient-internal replication of a single experiment. It is inappropriate to draw theoretical inferences from a single unreplicated experimental finding.

The third and optional stage of the procedure is a detailed analysis of the patient’s performance in certain tests, rather than just using overall scores. This can include the effects of relevant stimulus variables, of different error types, of consistency of performance on the same items across replications of the same material, and so on.

The theoretical utility of the single-case study needs to be considered in the context of two different types of model. The first are models of the global cognitive architecture in a given domain in which the components are functional subsystems (Posner 1978) that are anatomically isolable and so can be individually damaged by neurological disease. Such models can be characterized as having a modular functional architecture where the term ‘modular’ is used with a less specific referent than in Fodor’s (1983) sense. For such models, when a single subsystem is damaged, one will obtain a dissociation between impaired performance on tasks which require the subsystem and intact performance on tasks which do not require it. A standard move in cognitive neuropsychology is to treat the existence of a set of dissociations in a patient between tasks Y₁ to Y_i and task N₁ to N_j as evidence in favor of models where the pattern can arise from damage to one or more subsystems or connections, and against models where no such combination can produce the observed overall pattern of performance. This however depends on the following assumptions which are derived in part from those of Caramazza (1986) (see Shallice 1988 for a much more extensive discussion):

(a) Cognitive subsystems are qualitatively and quantitatively similar across individuals, at least by comparison with the effects of neurological disease.

(b) The task procedure, used by the patient, in the sense of the cognitive schemas that are controlling the overall operation of the collection of functional subsystems involved, does not differ markedly from that in use by normal subjects. This implies that the patient is not using a strategy unusual in normals, as, for instance, in letter-by-letter reading, which involves sounding out each letter in turn and then attempting to read by combining them into a word, as is found in the syndrome pure alexia.

(c) Reorganization of function following the lesion has had no more than a secondary effect in task performance. Clearly if the patient’s cognitive system is qualitatively different from a normal system in its organization of subsystems, it will not be possible to make inferences from the patient’s performance to the organization of the normal system.

(d) The tasks N₁ to N_j, on which performance is impaired, are not more demanding of some critical cognitive resource than tasks Y₁ to Y_i where performance is intact.

The most complex of these is the last, the pre-supposition that for most behavioral purposes the effects of damage to a subsystem can be ordered on a single dimension—resource. It has been argued by Glymour (1994) that the introduction of this concept makes cognitive neuropsychology a theoretically in-tractable enterprise, as far as the determination of the modular functional architecture using the observed overall pattern of performance of patients is concerned. Glymour argues that the degrees of freedom open to possible explanations would be greater than the constraints imposed by the observations of task performance. However, that analyses based on the resource concept can in practice be used to differentiate between alternative models has been shown in the concrete case of phonological output buffer disorder by Shallice et al. (2000).

The introduction of the resource concept allows one to infer that if a complementary set of dissociations is observed in a second patient producing a double dissociation, in at least one of the two cases, the pattern of Y₁ to Y_i contrasting with N₁ to N_j cannot be explained merely in terms of differential resource demands on a single subsystem being made across the set of tasks (see Shallice 1988, Chap. 10 for extended discussion). It should be noted, however, that these arguments are restricted to modular cognitive architectures. The existence of a double dissociation does not entail that the underlying system is modular; other possibilities include a continuum of processing space, within which different regions are damaged, over-lapping processing regions, and systems in which different levels or aspects of operation of a subsystem can be separately damaged, as for instance on-line processing and the capacity to learn (see Shallice 1988, Chap. 11). However one alternative which does seem to be ruled out is an equipotential distributed system (Bullanaria and Chater 1995, but see Plaut 1995).

Inferences in neuropsychology from impaired behavior to mechanism are not, however, restricted to inferences based on the overall pattern of performance and the global organization of a modular functional architecture (see McCloskey 2000). The second type of model is those specifying how particular subsystems operate. More detailed aspects of the patient’s performance can give more specific information about the operation of a processing system. Thus, in attentional dyslexia, patients show migrations of letters from one word to another (e.g., ‘flip snag’ to ‘snip flag’) (Saffran and Coslett 1996). This is compatible with damage to a mechanism which specifies over which parts of the visual field the results of letter-form level analysis are admitted for higher-level processing (see Mozer 1991 for more detailed discussion of the theoretical implications of the disorder).

2. The Multiple-Case Study And The Functional Syndrome

Any individual patient’s performance can be influenced by a variety of factors which reduce its relevance for inferences to normal processing. The patient may have an idiosyncratic functional or anatomical organization of their cognitive system, use an unusual strategy in some tests, or the disorder may even be influenced by psychiatric factors, so-called ‘functional overlay.’ One pragmatic way of partially guarding against these possibilities is to report two or more patients showing the same pattern of symptoms.

Second, if a dispute arises about the interpretation of empirical aspects of the report of an individual case, there is no way it can be resolved, other than by further investigation of the patient’s difficulties, or reanalysis of existing findings. In practice these approaches are only rarely undertaken. Therefore, the recourse tends to be to find alternative arguments on the basic theoretical position held, in which case the original findings become redundant. (An example of an un-resolvable theoretical question in the individual patient is arguments of whether evidence for syllabic structure in graphemic buffer disorder patient LB represents the effects of orthographic or phonological based processes; see Caramazza and Miceli 1990, Jonsdottir et al. 1996).

One way of making the neuropsychological data-base more solid is to establish functional syndromes, or characteristics held in common by all of a group of patients. However, a combination of characteristics can arise because of the anatomical proximity of different functional systems, producing so-called associated deficits. The concept of functional syndrome has therefore often been rejected metatheoretically on the grounds that a single counterexample can lead to the collapse of the putative functional syndrome. This is not necessarily the case. The functional syndrome may fractionate with both subvarieties being theoretically interpretable; one example is the two varieties of so-called letter-by-letter readers who circumvent the word-reading difficulty induced by their pure alexia. One group uses the names of the letter forms and knowledge of the spelling; the other uses a sounding-out strategy (see Patterson and Kay 1982). Moreover, the overall performance of a single patient can also derive from multiple impairments. This is not therefore an argument against the functional syndrome perspective, given that one restricts functional syndromes to ones in which all patients behave in the same fashion, and treats the subsyndromes resulting from fractionation of an original functional syndrome as each requiring theoretical explanation. The functional syndrome is therefore a concept which has the utility of producing a more solid database on which to assess alternative theories (see, for example, Shallice et al. 2000 for a concrete example of the approach).

Functional syndromes are of relevance for theories other than modular ones. Connectionist models can predict functional syndromes with complex characteristics too (see, for example, Mozer and Behrmann 1990). However, a different type of multiple single-case study is more usually relevant to this type of model. This is the use of a database for confronting theory of all patients showing deficits in a particular processing domain; their behavior is then individually fitted, but using different parameter values within a common underlying model. The prototypic example of this approach is the study of Dell et al. (1997) on naming errors in aphasics (but see Ruml et al. in press for detailed criticism).

3. The Anatomically Based Group Study

A method which is very frequently used is to group patients according to the location of lesion and average the results of each group. A potential problem with this approach is that patients in a given group may vary qualitatively as well as quantitatively. The reason is simple. The defining anatomical region will typically involve subsystems other than that which is the critical theoretical focus of the study. Thus Caramazza’s (1986) argument on the functional heterogeneity of the group becomes relevant, in particular that the pattern of performance of the average of the group may not correspond to that of any individual representative of the group.

This possibility should not, however, be thought to invalidate the approach completely. While it presents a theoretical danger, its potency depends upon the nature of the averaged data and the nature of this theory. Thus, for certain types of theories the same problem occurs when the findings of normal subjects are confronted with theory. Second, for certain types of evidence, in particular double dissociations, it requires the conditions, which are a priori highly unlikely, for the pattern to arise as an average artifact and not be present in individual complementary pairs of subjects within the contrasting groups (see Shallice 1988, Chap. 9). Thus the existence of the double dissociation in the group data virtually always entails it occurring in specific pairs of individual subjects. This means that the situation when making inferences from the overall pattern of performance of groups to theory is virtually equivalent to that when making inferences from individual subjects. I know of no actual case where theoretical inferences from an anatomically based group study have been criticized as being vitiated by an averaging artifact.

A further key issue in anatomically based group studies is how to define the relevant region for patient inclusion. Typically, regions are defined a priori, such as lesions being confined to or involving a particular lobe. However, approaches are being developed in which determination of group membership is in-directly influenced by statistical procedures for analyzing functional-imaging results, such as statistical parametric mapping (SPM) (Friston 1997). This has resulted in the development of more complex procedures for determining the specification of the anatomical regions most appropriate for most sharply differentiating good and poor performance on a given measure. For instance, Stuss et al. (1998) have used classification and regression test (CART) procedure (Breiman et al. 1984) to assign patients to subgroups.

4. The Functionally Based Group Study Approach

An alternative methodology is to constitute groups according to the functional characteristics, such as when studies contrast Broca’s aphasics and Wernicke’s aphasics, or amnesics and normal controls. In many respects the methodological issues raised are similar to those of the previous approach. However, a critical problem relates to the criteria used for group definition, and hence the possibility of replication. In studies contrasting amnesic and control groups, for instance, this has often been a critical issue (see Butters and Cermak 1974). However, in aphasia research it has been a key issue since the 1980s. Thus there has been a very long-standing and as yet unresolved controversy on whether the comprehension performance of Broca’s aphasics mimics their production problems, in the sense of showing a dissociation between comprehension of active and passive sentences. Some early cases seemed to indicate that in certain patients it did not (see Berndt et al. 1996).

However, Grodzinsky et al. (1999) reject this meta-analysis on the grounds that it applied too loose criteria for patient selection. Some patients included, it was argued, were not Broca’s aphasics. Caramazza et al. (2000) have pointed out that Berndt et al.’s findings are still obtained when the more restrictive criteria, endorsed by Grodzinsky et al., are applied.

To be satisfactory, this approach must be based on effective functional syndromes that do not fractionate and on good operational criteria for diagnosing them; Broca’s and Wernicke’s aphasia probably do not pass the test. The danger of the clinically based group study approach is that if the functional syndrome which is used to define group membership does fractionate, then the critique of Caramazza (1986) will be potentially highly relevant, and only very simple characteristics of the averaged findings will prove inferentially relevant.

Bibliography:

1. Baddeley A D 1986 Working Memory. Clarendon Press, Oxford, UK
2. Berndt R S, Mitchum C C, Haedinges A W 1996 Comprehension of reversible sentences in agrammatism: A meta-analysis. Cognition 58: 289–308
3. Brieman L, Friedman J H, Olshen R A, Store C J 1984 Classification and Regression Trees. Wadworth, Belmont, CA
4. Bullinaria J A, Chater N 1995 Connectionist modelling: Implications for cognitive neuropsychology. Language and Cognitive Processes 10: 227–64
5. Butters N, Cermak L S 1974 Some comments on Warrington and Baddeley’s report of normal short-term memory iamnesic patients. Neuropsychologia 12: 283–5
6. Caramazza A 1986 On drawing inferences about the structure of normal cognitive systems from the analysis of patterns of impaired performance: The case for single-patient studies. Brain and Cognition 5: 41–66
7. Caramazza A, Hillis A E 1990 Spatial representation of words in the brain implied by studies of a unilateral neglect patient. Nature 346: 267–9
8. Caramazza A, Miceli G 1990 The structure of graphemic representations. Cognition 27: 243–97
9. Caramazza A, Cappa E F, Ray A, Berndt R S 2001 Agrammatic Broca’s aphasia is not associated with a single pattern of comprehension performance. Brain and Language 76: 158–84
10. Caramazza A, Miceli G, Villa G, Romani C 1987 The role of the graphemic buffer in spelling: Evidence from a case of acquired dysgraphia. Cognition 26: 59–85
11. Dell G S, Schwartz M F, Martin N, Saffran E M, Gognon D A 1997 Lexical access in normal and aphasic speech. Psychological Review 104: 801–38
12. Ellis A W 1987 Imitations of modelarity, or, the modularity of mind: Doing cognitive neuropsychology without syndromes. In: Coltheart M, Sartori G, Job R (eds.) The Cognitive Neuropsychology of Language. Erlbaum, London
13. Farah M J, McClelland J L 1991 A computational model of semantic memory impairment: Modality specificity and emergent category specificity. Journal of Experimental Psychology General 120: 339–57
14. Fodor J A 1983 The Modularity of Mind. MIT Press, Cambridge, MA
15. Friston K 1997 Analysing brain images: Principles and overview. In: Frackowiak R S J et al. (eds.) Human Brain Function. Academic Press, San Diego, CA, pp. 25–42
16. Garrard P, Patterson K, Watson P C, Hodges J R 1998 Category specific semantic loss in dementia of Alzheimer’s type. Functional-anatomical correlations from cross-sectional analyses. Brain 121: 633–46
17. Glymour C 1994 On the methods of cognitive neuropsychology. British Journal of Philosophy of Science 45: 815–35
18. Gonnerman L, Anderson E S, Devlin J T, Kempler D, Seidenberg M S 1997 Double dissociation of semantic categories in Alzheimer’s disease. Brain and Language 57: 254–79
19. Goodglass H, Quadfasel F A, Timberlake W H 1964 Phrase length and the type and severity of aphasia. Cortex 1: 133–53
20. Grodzinsky Y, Pinango M M, Zurif E, Drai D 1999 The critical role of group studies in aphasia: Comprehension regularities in Broca’s aphasia. Brain and Language 67: 134–47
21. Head H 1926 Aphasia and Kindred Disorders of Speech. Cambridge University Press, Cambridge, UK
22. Jonsdottir M, Shallice T, Wise R 1996 Language-specific processes in graphemic buffer disorder. Cognition 59: 169–97
23. McCloskey M 2001 The future of cognitive neuropsychology. In: Rapp B (ed.) The Handbook of Cogniti e Neuropsychology. Psychology Press, Philadelphia, pp. 593–610
24. Milner B 1958 Psychological deficits produced by temporal-lobe excision. Research Publications, Association for Research in Nervous and Mental Disease 36: 244–57
25. Moore C J, Price C J 1999 A functional neuroimaging study of the variables that generate category-specific object processing diff Brain 122: 943–62
26. Mozer M C 1991 The Perception of Multiple Objects. MIT Press, Cambridge, MA
27. Mozer M C, Behrmann M 1990 On the interaction of selective attention and lexical knowledge: A connectionist account of neglect dyslexia. Journal of Cognitive Neuroscience 2: 96–123
28. Nielson J M 1946 Agnosia, Apraxia, Aphasia: Their Value in Cerebral Localization. Hoeber, New York
29. Patterson K E, Kay J 1982 Letter-by-letter reading: Psycho-logical descriptions of a neurological syndrome. Quarterly Journal of Experimental Psychology, Series A 34: 411–41
30. Plaut D C 1995 Double dissociation without modularity: Evidence from connectionist neuropsychology. Journal of Clinical and Experimental Psychology 17: 291–321
31. Plaut D C, Shallice T 1993 Deep dyslexia: A case study of connectionist neuropsychology. Cognitive Neuropsychology 10: 377–500
32. Posner M I 1978 Chronometric Explorations of Mind. Erlbaum, Hillsdale, NJ
33. Ruml W, Caramazza A 2000 An evaluation of a computational model of lexical access: Comment on Dell et al. 1997 Psychological Review 107: 609–34
34. Saffran E M, Coslett B 1996 Attentional dyslexia in Alzheimer’s disease: A case study. Cognitive Neuropsychlogy 13: 205–28
35. Shallice T 1988 From Neuropsychology to Mental Structure. Cambridge University Press, Cambridge, UK
36. Shallice T, Warrington E K 1970 Independent functioning of the verbal memory stores: A neuropsychological study. Quarterly Journal of Experimental Psychology 22: 261–73
37. Shallice T, Rumiati R I, Zadini A 2000 The selective impairment of the phonological output buff Cognitive Neuropsychology 17: 517–46
38. Stuss D T, Benson D F 1984 Neuropsychological studies of the frontal lobes. Psychological Bulletin 95: 3–28
39. Stuss D T, Alexander M P, Hamer L, Palumbo C, Dempster R, Binns M, Levine B, Izuakawa D 1998 The effects of focal anterior and posterior brain lesions on verbal fluency. Journal of the International Neuropsychological Society 4: 265–78
40. Warrington E K, McCarthy R 1983 Category specific access dysphasia. Brain 106: 859–78
41. Warrington E K, Shallice T 1984 Category specific semantic memory impairment. Brain 107: 829–54