Cognitive neuropsychology attempts to reveal the patterns of cognition in the brain by the means of studying patients with abnormal brain functions (Plaut, 1995). Within the field of cognitive neuropsychology the prevalent hypothesis is that brain is composed of relatively independent components, or modules, and that these subsystems perform particular functions, which can due to brain damage be selectively injured (Plaut, 1995).
Alternative nonmodular theories emphasize the brain to be operating ‘as a whole’, like a ‘uniform general purpose system’ (Bechtel & Graham, 1998:p. 632). Between the debates of modularity and homogeneity, neuropsychologists have got a powerful counter argument: the occurrence of double dissociations, which are traditionally interpreted as offering strong evidence about the modularity of mind, and therefore they are the essence of theorizing in cognitive neuropsychology (Detre, 2000; Plaut, 1995).
A double dissociation is manifested when patients suffer brain impairment with the reversed pattern of deficits from each other. That is if one patient has dysfunctional cognitive ability A, but spared ability in cognitive function B, then the other patient demonstrates the opposite pattern with inability to perform cognitive function B and at least a relatively intact ability to perform function A (Ashcraft, 1994).
The underlying logic here is that these cognitive functions are psychologically and anatomically separate and distinct, because if they were not, the brain damage should then disrupt both functions equally (Ashcraft, 1994). We shall now continue to explain the concept of a double dissociation with a concrete demonstration of the phenomenon. Thus we shall contrast two types of aphasias (disruption of language): Broca’s and Wernicke’s aphasias, as they together offer an excellent example of a double dissociation.
We will also examine why information gained through studying a single dissociation is inappropriate on its own – that is without a double dissociation study- in providing valid evidence about the modularity of mind. We shall then continue by pointing out some critical aspects within the methodology of lesion studies, followed by information about double dissociations in the light of a connectionist model, which assumes the brain to be operating in a parallel distributed process, in which a lot of simple units cooperate in order to yield complex activities (Ashcraft, 1994).
The point here is that even though the relevance of the connectionist model in explaining human cognition remains uncertain, we will see that this nonmodular system has generated double dissociations, which challenges the traditional assumptions about modularity (Bechtel & Graham, 1998). A single dissociation occurs when a patient performs at least relatively normally on one task and very poorly on another task (Shallice, 1988, cited in Bechtel & Graham, 1998).
Let us illustrate this by using Broca’s aphasia as an example: Broca’s aphasia (or non fluent aphasia) is typically characterized by disrupted speech production, disrupted syntax, but spared semantics (Detre, 2000). Hence Broca’s aphasics are able to understand language, but their speech is extremely clumsy, mainly consisting of two word utterances with only a few if any reference to grammar (Ashcraft, 1994; Gleitman et al. , 1999).
Broca’s area is located in the anterior lateral portion of the left hemisphere of the brain, which lies adjacent to the motor projection areas related to speech, thus explaining the disrupted motor word production in Broca’s aphasia (Buchel, Frith, Friston, 1999; Gleitman et al. , 1999). Impairment in motor word production cannot be induced by brain damage in any other area, which seems to offer evidence that the cognitive function is localized (Ashcraft, 1994; Detre, 2000).
However, there are other possible explanations as well. One could argue, that Broca’s area, rather than being independent, is a part of a hierarchical system of relationships, and damage to this area contributes other parts in the systems to receive insufficient or wrong inputs, leading the whole system to fail in a particular cognitive task (in this case the production of speech) (Bechtel & Graham, 1998; Detre, 2000). Moreover, it could be that with a single dissociation one task is simply harder than another one.
For example, it could be that the lesion has damaged both cognitive representations equally, such as speech production and comprehension, but it is more difficult for a patient to produce speech than to comprehend it (Plaut, 1995). However, if a double dissociation is manifested, it undermines the basis for these claims. The symptoms of Wernicke’s aphasia (or fluent aphasia) are almost like a mirror reversion of the complaints associated with Broca’s aphasia, and this pattern of complimentary deficits is precisely the basis for a double dissociation.
Unlike in Broca’s aphasia, in Wernicke’s aphasia language comprehension and semantics are disrupted while as speech production is spared (Detre, 2000): The flow of speech is effortless for a patient with Wernicke’s aphasia, but s/he is unable to retrieve the right word and consequently s/he produces nonsensical ‘wordsalads’ with very little meaning (Gleitman et al. , 1999). Wernicke’s aphasia may be caused by a lesion in the auditory primary projection area of the left temporal and parietal lobes, which has been interpreted to affect sensory word representations, hence giving rise to the disrupted comprehension (Gleitman et al. 1999; Martin, 2003). Thus, through impairment in different areas of the brain, patients show the opposite pattern of deficits to each other. In this way the double dissociation cancels out the basis of the claims that the test results could be explained in terms of task difficulty, and Broca’s and Wernicke’s aphasics provide strong evidence that speech production and comprehension are separately functioning components (Detre, 2000). However, Martin (2003) emphasizes that these traditional views of language representations are oversimplified.
The suggestion that motor word representations are localized in Broca’s area and auditory words in Wernicke’s area, do not provide a valid understanding of the underlying reasons in language syndromes, and cannot capture the complicated nature of language representation in the human brain (Martin, 2003). For example, it has been argued that there are a number of possible underlying reasons that can give rise to the poor comprehension typical for Wernicke’s aphasics, including a disruption of phoneme identification, a disruption of phonological word forms, or a disruption of semantics (Martin, 2003).
Martin (2003) also points out that studies during the 1970s ; 1980s revealed that some Broca’s aphasics showed poor comprehension when testing their understanding of a complex syntactic information in a sentence, which obviously cannot be explained in terms of a disrupted motor word production. This uncovers the very heart of the problems in double dissociation studies: It is extremely difficult to find patients with districted impairments; patients who have a lesion that is limited into a single component. Instead, brain damage is usually extensive, affecting different cognitive systems to a greater or lesser extent (Plaut, 1995).
Therefore, rather than being dependent on finding patients with districted impairments, it could be more appropriate to concentrate on individual cases, engaging into a theory based functional analysis, and through this type of systematic research we may gain a fuller understanding of the overall effects caused by brain damage (Plaut, 1995; Martin, 2003). Plaut (1995) criticizes that many authors seem to unquestionably interpret double dissociations as implying the existence of independent modules, because it is ‘the simplest and so, intuitively, the most natural interpretation’ (Plaut, 1995, cited in Bechtel & Graham, 1998: 633).
Shallice (1988) points out that modularity seems to be a reasonably valid interpretation of the occurrence of double dissociations, but only so, if it is the only kind of system that is able to give rise to double dissociations. Shallice continues to undermine the theory of modularity by proposing alternative nonmodular, or partially modular systems that could potentially evoke double dissociations (Bechtel & Graham, 1998; Plaut, 1995).
Moreover, Plaut & Shallice (1993) found that nonmodular connectionist network generated a double dissociation between concrete and abstract word reading after damage to a connectionist network that pronounces words via meaning without having separable components (Plaut, 1995). Martin (2003) also points out that many studies have manifested double dissociations in the production of function word versus content words, for example between nouns and verbs. These findings suggest that in some level of word production, the grammatical aspect of a sentence is distinguished neurally (Martin, 2003).
We have now identified the two prevalent 20th century paradigms within cognitive neuropsychology: that of functional segregation and nonmodular models, which emphasize independent components or homogeneity respectively. Through the lack of available resources it was extremely difficult to further develop the nonmodular approaches to the functioning of the brain. Hence, while nonmodular models stood in the shadow, the dominant paradigm has emphasized the functional segregation of the brain, being strongly reinforced by the evidence from double dissociation reports (Buchel et al. 1999). However, -as briefly addressed in the discussion about Broca’s and Wernicke’s aphasia- only recently, through more detailed lesion studies, we have become fully aware that the relation between lesion site and aphasic deficit is far from perfect (Buchel et al. , 1999). Moreover, connectionist neuropsychology has greatly challenged traditional assumptions about modularity by reporting double dissociations produced by lesioning a connectionist network that has no distinct modules (Plaut, 1995).
These deficits provide evidence towards functional specialisation that is not ultimately related to the network’s structure (Plaut, 1995; Martin, 2003). Martin (2003) writes that in order to report whether consistent localisation of language in the human brain can be identified, it would be better to study individual cases of patients with brain damage with theory based analysis, instead of using group comparisons. There is some variety within the modularist accounts of how modules should be defined, but it is certainly clear that the brain is not a combination of completely independent small units (Buchel et al. 1999; Plaut, 1995).
Instead the connectivity within the brain allows components to be spread over a wide area, with the result that we are not necessarily able to pinpoint the exact location of modules that are responsible for particular sub-functions (Detre, 2000). Ashcraft (1994) writes that the distinction between modular and nonmodular approaches could ‘turn out to be trivial or fundamental’ in the future (p. 511). He continues by speculating that ‘maybe future neural net models will need separate modules, or maybe the modularity of human cognition does not require modularity at the level of neural net’ (Ashcraft, 1994: 511).
The lesion method is not particularly good when studying interactions between brain regions (Buchel et al. , 1999). However, through the availability of new functional brain imaging techniques, the field of cognitive neuropsychology is in the middle of a fascinating period, which hopefully leads into a greater insight of the patterns of cognition in the human brain, thus perhaps resolving some of the question marks between modular and nonmodular theories (Pinel, 2002).