Alzheimer's disease (AD) is a neurodegenerative condition that results in insidiously progressive cognitive decline. According to widely recognized clinical diagnostic criteria for AD (McKhann et al., 1984), these patients have language processing impairments as well as difficulty with memory, visual perceptual-spatial processing, and executive functioning. The language impairment changes as the disease progresses (Bayles et al., 2000), and one profound consequence of language difficulty in AD is that this deficit strongly reflects clinical decline and the need for additional skilled nursing support (Chan et al., 1995). This article briefly summarizes the studies showing that AD patients' language deficit includes difficulty with comprehension and expression of the sounds/letters, words, and sentences that are used to communicate in day-to-day circumstances. Work relating these language deficits to a specific neuroanatomical distribution of disease is also reviewed.

We may consider first the semantic impairment in AD. This deficit limits the comprehension and expression of concepts represented by single words and sentences. In expression, for example, a significant word-finding deficit is a prominent and early clinical feature of AD (Bayles, Tomoeda, and Trosset, 1990; White-Devine et al., 1996; Cappa et al., 1998). This is seen in spontaneous speech as well as on measures of confrontation naming. Naming difficulty due to a semantic impairment often is manifested as semantic paraphasic errors, such as the substitution of “chair” for the intended target “table.” As the condition progresses, AD patients' spontaneous speech becomes limited to the use of overlearned phrases and ultimately becomes quite empty of content, while they fail to provide any responses during confrontation naming.

Semantic deficits are also prominent in comprehension. More than 50% of AD patients differ significantly from healthy seniors in their performance on simple category judgment tasks. For example, many AD patients are impaired when shown a word or a picture and asked, “Is this a vegetable?” (Grossman et al., 1996). Priming is a relatively automatic measure of semantic integrity: AD patients have relatively preserved priming for high-frequency lexical associates such as “cottage-cheese” that have little semantic relationship (Nebes, 1989; Ober et al., 1991), but they are impaired in their priming for coordinates taken from the same semantic category, such as “peach-banana” (Glosser and Friedman, 1991; Glosser et al., 1998). Item-by-item analyses show reduced priming for words that are difficult to understand and name (Chertkow, Bub, and Seidenberg, 1989). The unity of impairment across comprehension and expression is emphasized by the observation of the greatest naming difficulty in patients with significant semantic comprehension deficits (Chertkow and Bub, 1990; Hodges, Salmon, and Butters, 1992). The basis for this pattern of impaired semantic memory has been an active focus of investigation. Some studies associate the semantic comprehension impairment in AD with the degradation of knowledge about a word and its associated concept (Gonnerman et al., 1997; Garrard et al., 1998; Conley, Burgess, and Glosser, 2001). A category-specific deficit understanding or naming natural kinds such as “animals” compared with manufactured artifacts such as “implements” may emerge in AD (Silveri et al., 1991). Other recent work suggests that difficulty understanding words and pictures in AD is related to an impairment in the categorization process that is so crucial to understanding concepts. In particular, AD patients appear to have difficulty implementing rule-based processes for understanding the critical features of words that determine category membership (Grossman, Smith, et al., submitted) or for learning the category membership of new concepts (Koenig et al., 2001).

The comprehension and expression of concepts often requires appreciating the long-distance relationships among several words in a sentence. Some early work attributes sentence processing difficulty in AD to a grammatical deficit. Paragrammatic errors such as “mices” and “catched” can be observed in speech, oral reading, and writing. Other studies relate impaired comprehension to difficulty with the grammatical features of phrases and a deficit in understanding grammatically complex sentences such as those containing a center-embedded clause (Emery and Breslau, 1989; Kontiola et al., 1990; Grober and Bang, 1995; Croot, Hodges, and Patterson, 1999). Essentially normal performance during on-line studies of sentence comprehension cast doubt on this claim (Kempler et al., 1998; Grossman and Rhee, 2001). More recently, considerable evidence indicates that sentence processing difficulty is related to a limitation in the working memory resources often needed to support sentence processing. Although the precise nature of the limited cognitive resource(s) remains to be established, AD patients' grammatical comprehension deficit can be brought out by experimental manipulations that stress cognitive resources such as working memory, inhibitory control, and information processing speed. Studies demonstrate working memory limitations through the use of verbs featuring unusual syntactic-thematic mapping in a sentence comprehension task (Grossman and White-Devine, 1998), concurrent performance of a secondary task during sentence comprehension (Waters, Caplan, and Rochon, 1995; Waters, Rochon, and Caplan, 1998), and limited inhibition of the context-inappropriate meaning of a polysemous word (Faust et al., 1997).

Semantic memory and sentence processing appear to be preserved in some AD patients. Nevertheless, this cohort of AD patients may have a different language impairment profile. Many of these patients, despite preserved single-word and sentence comprehension, are impaired in retrieving words from the mental lexicon. This kind of naming difficulty is marked by changes in the sounds contributing to a word, such as omissions and substitutions (Biassou et al., 1995). This limitation in lexical retrieval appears to be equally evident on oral lexical retrieval tasks and in writing. AD patients appear to be quite accurate at discriminating between speech sounds that vary in the place of articulation or voice onset timing, but phonemic (single-sound) substitutions can also be heard in their speech.

Alzheimer's disease is a focal neurodegenerative condition. Functional neuroimaging studies obtained at rest with modalities such as single-photon emission computed tomography, positron emission tomography, and functional magnetic resonance imaging (fMRI) (Foster et al., 1983; Friedland, Brun, and Budinger, 1985; DeKosky et al., 1990; Johnson et al., 1993; Alsop, Detre, and Grossman, 2000) and histopathological studies of autopsied brains (Brun and Gustafson, 1976; Arnold et al., 1991; Braak and Braak, 1995) show that specific brain regions are compromised in AD. The neuroanatomical distribution of disease revealed by these studies includes gross defects such as atrophy and microscopic abnormalities such as neuritic plaques and neurofibrillary tangles in the temporal, parietal, and frontal association cortices of the brain. The neural basis for the language difficulties in AD is investigated most commonly through brain-behavior correlation studies, although occasional functional neuroimaging reports describe defects in regional brain activation during language challenges. Early correlation studies relate sentence comprehension difficulty to reduced resting activity in posterior temporal and inferior parietal regions of the left hemisphere (Haxby et al., 1985; Grady et al., 1988). More recent work associates difficulty understanding single words and impaired confrontation naming with left temporoparietal cortex (Desgranges et al., 1998). Moreover, the defect in this brain region is significantly greater in AD patients with a semantic memory impairment than in AD patients with relatively preserved semantic memory (Grossman et al., 1997), and a comparative study demonstrates the specificity of this correlative pattern in AD relative to patients with a frontotemporal form of dementia (Grossman et al., 1998). By comparison, only very modest correlations show a relationship between grammatical comprehension and left inferior frontal cortex in AD.

A handful of functional neuroimaging studies report monitoring the regional cortical responses of AD patients during language challenges. One study shows limited activation of middle and inferior frontal regions in AD patients that had been recruited during a category membership semantic decision in healthy seniors (Saykin et al., 1999). More recently, a BOLD fMRI study of semantic judgments described limited activation of left temporoparietal cortex and frontal cortex in AD patients compared to healthy seniors, and AD patients recruited brain regions adjacent to the activated areas seen in elderly control subjects for specific categories of knowledge such as “animals” and “implements” (Grossman, Koenig, et al., in press).

AD patients thus have prominent deficits at several levels of language processing. This includes impaired semantic memory, manifested in measures of comprehension and expression. There is also difficulty with lexical retrieval in reading and writing, although perceptual judgments of speech sounds are relatively preserved. The neural basis for these language impairments appears to be a defect in temporoparietal association cortex of the left hemisphere, although a defect in left frontal association cortex also may contribute to the language impairments in AD.

See also dementia.