Connectionist models of aphasic syndromes first emerged in the late nineteenth century. Broca (1861) described a patient, Leborgne, whose speech was limited to the monosyllable tan but whose ability to understand spoken language and nonverbal cues and ability to express himself through gestures and facial expressions were normal. Leborgne's brain contained a lesion whose center was in the posterior portion of the inferior frontal convolution of the left hemisphere, now known as Broca's area. Broca claimed that Leborgne had lost “the faculty of articulate speech” and that this brain region was the neural site of the mechanism involved in speech production. In 1874, Karl Wernicke described a patient with a speech disturbance that was very different from that seen in Leborgne. Wernicke's patient was fluent, but his speech contained words with sound errors, other errors of word forms, and words that were semantically inappropriate. Unlike Leborgne, Wernicke's patient did not understand spoken language. Wernicke related the two impairments—one of speech production and one of comprehension—by arguing that the patient had sustained damage to “the storehouse of auditory word forms,” leading to speech containing the types of errors that were seen and impaired comprehension. By extrapolation from a similar case he had not personally examined, Wernicke concluded that the patient's lesion was in the posterior portion of the left superior temporal gyrus, now known as Wernicke's area, and that this region was the locus of the “storehouse of auditory word forms.” Wernicke argued that, in speaking, word sounds were conveyed from Wernicke's area to Broca's area, where the motor programs for speech were developed. This connection gave this type of model its name.

Lichtheim (1885) deveoped a more general model of this type. Lichtheim recognized seven syndromes, listed in Table 1. Lichtheim argued that these syndromes followed lesions in the regions of the brain depicted in Figure 1. These syndromes were criticized on neuroanatomical grounds (Marie, 1906; Moutier, 1908), dismissed as simplifications of reality that were of help only to schoolboys (Head, 1926), and ignored in favor of different approaches to language (Jackson, 1878; Goldstein, 1948). Nonetheless, they endured. Benson and Geschwind (1971) reviewed the major approaches to aphasia as they saw them and concluded that all researchers recognized the same basic patterns of aphasic impairments, despite using different nomenclature. Three more syndromes have been added by theorists such as Benson (1979), and Lichtheim's model has been rounded out with specific hypotheses about the neuroanatomical bases for several functions that he could only guess at.

Figure 1..  

The classical connectionist model (modified from Lichtheim, 1885). W indicates Wernicke's area, the site of long- term storage of word sounds. B indicates Broca's area, the site for speech planning. C represents the concept center, which Wernicke thought was diffusely located in parietal lobe. Information flows along the pathways indicated by lines. The presence of these pathways (“connections”) gives this type of model its name.

Additional neuroanatomical foundation was first suggested in a very influential paper by Geschwind (1965). Geschwind argued that the inferior parietal lobe was a tertiary association cortical area that received projections from the association cortex immediately adjacent to the primary visual, auditory, and somesthetic cortices in the occipital, temporal, and parietal lobes. Because of these anatomical connections, the inferior parietal lobe served as a cross-modal association region, associating word sounds with the sensory qualities of objects. This underlay word meaning, in Geschwind's view. Damasio and Tranel (1993) extended this model to actions, arguing that associations between word sounds and memories of actions were created in the association cortex in the inferior frontal lobe. Geschwind (1965) and Damasio and Damasio (1980) also argued that the anatomical link between Wernicke's area and Broca's area (in which a lesion caused conduction aphasia) was the white matter tract known as the arcuate fasciculus.

The three and a half decades that have passed since publication of Geschwind's paper have brought new evidence for these syndromes and their relationships to brain lesions. Aphasic syndromes have been related to the brain using a series of neuroimaging techniques, first radionuclide scintigraphy with technetium⁹⁹, then computed tomography, magnetic resonance imaging, and positron emission tomography. All have confirmed the relationship of the major syndromes to lesion locations. These aphasic syndromes and their relationships to the brain figure prominently in recent reviews of aphasia in leading medical journals (e.g., Damasio, 1992).

Despite this revival, the connectionist approach to aphasic syndromes is under renewed attack.

A major limitation of the classical syndromes is that they stay at arm's length from the linguistic details of language impairments. The classical aphasic syndromes basically reflect the relative ability of patients to perform entire language tasks (speaking, comprehension, etc.), not the integrity of specific operations within the language processing system. Linguistic descriptions in these syndromes are incomplete and unsystematic. For instance, the speech production problem seen in Broca's aphasia can consist of one or more of a large number of impairments: dysprosodic speech, poorly articulated speech, agrammatism, an unusual number of short phrases. If all we know about a patient is that she or he has Broca's aphasia, we cannot tell which of these problems (or other) that person has.

A second problem is that identical deficits occur in different syndromes. For instance, certain types of naming problems can occur in any aphasic syndrome (Benson, 1979). Because of this, most applications of the clinical taxonomy result in widespread disagreements as to a patient's classification (Holland, Fromm, and Swindell, 1986) and to a large number of “mixed” or “unclassifiable” cases (Lecours, Lhermitte, and Bryans, 1983). The criteria for inclusion in a syndrome are often somewhat arbitrary: How bad does a patient's comprehension have to be for the patient to be identified as having Wernicke's aphasia instead of conduction aphasia, or global aphasia instead of Broca's aphasia? There have been many efforts to answer this question (see, e.g., Goodglass and Kaplan, 1972, 1982; Kertesz, 1979), but none is satisfactory.

A third problem with the classical aphasic syndromes is that they are not as well correlated with lesion sites as the theory claims they should be. These syndromes are related to lesion sites reasonably well only in cases of rapidly developing lesions, such as stroke. Even in these types of lesions, the syndromes are never applied to acute and subacute phases of the illness. Even in the chronic phase of diseases such as stroke, at least 15% of patients have lesions that are not predictable from their syndromes (Basso et al., 1985), and some researchers think this figure is much higher—as much as 40% or more, depending on what counts as an exception to the rule (de Bleser, 1988). We now know that the relationship between lesion location and syndrome is more complex than we had thought, even in cases in which the classical localization captures part of the picture. Broca's aphasia, for instance, does not usually occur in the chronic state after lesions restricted to Broca's area but requires much larger lesions (Mohr et al., 1978). Some theorists have argued that the localizing value of the classical syndromes reflects the co-occurrence of variable combinations of language processing deficits with motor impairments that affect the fluency of speech (Caplan, 1987; McNeil and Kent, 1991). From this point of view, the localizing value of the classical syndromes is due to the invariant location of the motor system.

Finally, the classical syndromes offer very limited help to the clinician planning therapy, because the syndromes give insufficient information about what is wrong with a patient. For example, knowing that a patient has Broca's aphasia does not tell the therapist what aspects of speech need remediation—articulation of sound segments, prosody, production of grammatical elements, formulation of syntactic structures, and so on. Nor does it guarantee that the patient does not need therapy for a comprehension problem; it only implies that any comprehension problem is mild relative to the problems of other aphasics or to the patient's speech problem. Finally, it does not guarantee that the patient does not have other problems, such as anomia, difficulty reading, and the like. In practice, most clinicians do not believe that they have adequately described a patient's language problems when they have identified that patient as having one of the classic aphasic syndromes. Rather, they specify the nature of the disturbance found in the patient within each language-related task; for example, they indicate that a patient with Broca's aphasia is agrammatic, has a mild anomia, and so on. Detailed psycholinguistic and linguistic descriptions of aphasic impairments are slowly replacing the disconnection approach to syndromes.

It is a feature of the history of science, and some think a tenet of the philosophy of science, that people do not abandon a theory because it has inadequacies. Some philosophers of science think that no theory is ever proven wrong. According to this view, theories are abandoned because people get tired of them, and people get tired of theories because they have others that they think are better. This perspective on science applies to the connectionist approach to aphasic syndromes. The classic syndromes have not been abandoned, but their acceptance is waning, and there are new developments that address some of their inadequacies.