One of the more active areas of clinical research related to dementia and Alzheimer’s disease is the attempt to identify cognitive, behavioral, and biological changes that occur during a “preclinical” phase that precedes the manifestation of the overt dementia syndrome. This research is based on current conceptualizations of Alzheimer’s disease, which suggest that the neuropathology of the disease gradually accumulates, causing the insidious onset of cognitive dysfunction. As neuropathological changes proceed, cognitive dysfunction increases to a level where it becomes unambiguous and disruptive to daily functioning. It is at this point that the clinical diagnoses of dementia and Alzheimer’s disease can be made with some certainty. Studies examining cognitive changes in the preclinical phase of Alzheimer’s disease usually identify decline in episodic memory (the contextually rich memory of specific events and places a person experiences) as the earliest manifestation of the disease. Memory decline in the face of relatively intact general cognition (e.g., as assessed with a brief mental status examination) and day-to-day functioning defines a condition widely known as amnestic mild cognitive impairment.

A growing body of evidence based on extensive and detailed neuropsychological testing suggests that cognitive changes in preclinical Alzheimer’s disease are more global than once thought (1). One domain of cognition that appears to be adversely affected in patients with mild cognitive impairment is semantic memory. Semantic memory refers to our general fund of knowledge, which consists of the meanings and representations of words, concepts, and overlearned facts that are not dependent on contextual cues for their retrieval. This knowledge is usually assumed to be organized as a complex associative network in which concepts that have many attributes in common are more strongly associated than those that share fewer attributes. Semantic knowledge is thought to be stored in a distributed manner in neocortical association areas of the temporal and parietal lobes and is not dependent on the medial temporal lobe structures that are important for episodic memory. Semantic memory is often impaired early in the course of Alzheimer’s disease, as shown by patients’ reduced ability to recall overlearned facts (e.g., the number of days in a year) and their impairment on tests of confrontation naming and verbal fluency (2). The impairment is thought to reflect the loss of semantic knowledge, rather than inefficient retrieval, because studies that have probed for knowledge of particular concepts across different modes of access and output (e.g., fluency, confrontation naming, sorting, word-to-picture matching, and definition generation) demonstrate that patients with Alzheimer’s disease are significantly impaired across all tasks, and there is item-to-item correspondence so that when a particular stimulus item is missed (or correctly identified) in one task, it is likely to be missed (or correctly identified) in other tasks that access the same information in a different way (3).

A growing body of evidence suggests that subtle deficits in semantic memory occur in patients with amnestic mild cognitive impairment and may be an indication of imminent progression to dementia (4). Patients with amnestic mild cognitive impairment are impaired on semantically demanding naming tasks that require producing proper nouns, such as the names of famous people or buildings (4, 5). They are also often impaired when required to generate exemplars from a specific semantic category (e.g., animals) but not when required to rapidly generate words beginning with a particular letter (e.g., F, A, or S) (5, 6). This pattern of performance is also present in mildly demented patients with Alzheimer’s disease and is often interpreted as a reflection of semantic knowledge loss because the semantic fluency task places greater demands than the letter fluency task on the use of semantic organization to efficiently generate words from a small and highly related set of exemplars.

In this issue of the Journal, Kirchberg et al. (7) report a study that compared the integrity of semantic memory in 25 patients with amnestic mild cognitive impairment, 27 mildly demented patients with Alzheimer’s disease, and 70 healthy comparison subjects. Semantic memory was assessed with a semantic distance task in which participants had to judge which of two entities, presented as images or words, was larger in the real world. Semantic distance was manipulated by adjusting the disparity in real-world size between the two entities that were being compared. Entities similar in real-world size have a small semantic distance (e.g., a key versus an ant), whereas those with a large size disparity have a large semantic distance (e.g., a house versus an ant). Images were presented in two conditions: congruent, in which the image of the larger real-world object was larger than the image of the smaller object (e.g., a house pictured larger than an ant), and incongruent, in which the image of the smaller object was larger than the image of the larger object (e.g., an ant pictured larger than a house). Results revealed that the Alzheimer’s disease and mild cognitive impairment groups were less accurate and slower than the healthy comparison group in making semantic decisions for word stimuli, and this deficit worsened as the sizes of the compared objects became more similar. The same result was found for images when collapsed across congruent and incongruent conditions, but only for accuracy. The image effect was largely driven by the incongruent condition in which patients with mild cognitive impairment performed similarly to healthy subjects in tasks with large semantic distances and similarly to Alzheimer’s disease patients in tasks with the more semantically demanding small distances. The semantic memory deficit demonstrated by this study adds confidence to the growing perception that subtle decline in this cognitive domain occurs in patients with amnestic mild cognitive impairment. Because the task places minimal demands on the effortful retrieval process, overt word retrieval, or language production, it also suggests that this deficit reflects an early and gradual loss of integrity of semantic knowledge.

A second important aspect of this study is the demonstration that semantic memory decrements in patients with mild cognitive impairment may contribute to a decline in the ability to perform usual activities of daily living. When combined with a measure of delayed recall, accuracy with large and small semantic distances and reaction time with large semantic distances in the congruent image condition accounted for 35% of the variance in scores obtained on a performance-based measure of functional abilities that used proxy tasks and scenarios, such as writing a check, dialing a telephone number, or planning a trip. This result indicates that a detectable decline in daily function occurs in patients with mild cognitive impairment and suggests that loss of semantic knowledge may be an important reason for this decline. In addition to influencing day-to-day functioning, early loss of semantic knowledge may contribute to a decline in judgment as important concepts and choices that must be understood and weighed become less clear to the patient. In the area of financial decision making, for example, loss of understanding of the value of assets or the meaning of contractual terms could leave patients with mild cognitive impairment vulnerable to abuse or fraud.

Advances in the neuropsychological detection of preclinical Alzheimer’s disease have occurred in parallel with advances in detecting reliable biological markers of the disease. It is now possible to use MRI to detect reductions in hippocampal volume and cortical thickness typically associated with Alzheimer’s disease, to use positron emission tomography imaging with amyloid binding to detect the deposition of β-amyloid in the brain, and to use biochemical assays of CSF to detect abnormal levels of the β-amyloid and tau proteins that constitute the plaques and tangles of Alzheimer’s disease (8). Each of these biomarkers is effective at predicting the development of Alzheimer’s dementia in nondemented elderly individuals. These discoveries have prompted a revision of the research diagnostic criteria for Alzheimer’s disease that have been widely used for more than 30 years. The revised criteria continue to define Alzheimer’s dementia largely as before (9) but now also incorporate the presence of a biomarker as supporting evidence. The intermediate stage of mild cognitive impairment that precedes frank dementia is now considered early Alzheimer’s disease, with varying degrees of confidence determined by the presence of biomarkers of the disease (10). Finally, an even earlier stage of preclinical Alzheimer’s disease is identified, which is characterized by the presence of biomarkers in asymptomatic individuals (11). Currently, the recommended use of biomarkers to detect early Alzheimer’s disease is limited to research, and thus the early clinical diagnosis continues to depend heavily on neuropsychological assessment to provide reliable symptom markers of the disease. The study by Kirchberg et al. identifies loss of integrity of semantic memory as one of these potential markers.