Oscar Wilde called memory "the diary that we all carry with us." His comparison of memory to a diary was intuitive because a diary is an organized store of specific information. Similarly, memory is defined as a composite of our stored representations of knowledge and the varied processes used to create and manipulate these representations. In the last two decades, there has been a revolution in our understanding of human memory. Human memory is no longer viewed as a unitary construct but as multidimensional and comprising several interrelated systems. Information about memory systems and how dementia affects them can be helpful to speech-language pathologists working with older adults.

Dementia is a syndrome characterized by acquired, persistent impairment of multiple cognitive domains—memory, language, attention, executive function, and visuospatial ability—severe enough to impair competence in daily living, occupation, and social interaction (Grabowski & Damasio, 2004).

Multiple Memory Systems

A thorough understanding of spared and impaired memory systems in dementia is vital for selecting and developing effective assessments and interventions. Humans have multiple memory systems—sensory, working, declarative, and nondeclarative—each of which depends on the integrity of different neuroanatomical structures (see Table 1 online [PDF]).

Figure 1 [PDF] introduces a contemporary model of human memory systems. Incoming stimuli first enter sensory memory, an extremely brief (1–2 seconds), modality-specific store of information received by the sensory receptors (eyes, ears).

Short-Term Memory

Stimuli then enter short-term memory, a concept Baddeley and Hitch (1974) suggested replacing with that of working memory (WM), a dynamic, short-term, limited-capacity buffer that enables storing information briefly and actively manipulating it while it is being processed. Baddeley described WM as a system with multiple components—a central executive, the phonological loop, the visuospatial sketchpad, and more recently, the episodic buffer. The central executive is a supervisory system that coordinates the other WM components, which are content-specific, short-term stores that allow recirculation of recently acquired information. When someone looks up a number in the phone book and repeats it to himself while dialing it, the phonological loop is engaged. The visuospatial sketchpad stores visual and spatial information briefly, such as object shapes and colors or a route within a building. The episodic buffer connects information across domains (verbal, visual, spatial) and helps chronologically order information (for example, memory of sequential events in a story).

Long-Term Memory

Information is stored permanently in long-term memory, which includes two components: declarative or explicit memory ("knows that") and nondeclarative (NDM) or implicit memory ("knows how").

Declarative memory includes: 

• Episodic memory (EM), which allows conscious recollection of specific episodes. EM is highly vulnerable to the effects of aging (Craik, 2000) and neurodegenerative disorders such as Alzheimer's disease.
• Semantic memory (SM), which is the extensive network of associations and concepts that form our knowledge of the world. In SM, when one concept, word, or event is activated, related concepts, events, and words also get activated (for example, activating the concept of Thanksgiving also activates associated concepts of November, family reunions, and turkey). 
• Lexical memory (LM), which is a distinct store of specialized word knowledge (Schank, 1975) organized as a network with words as concepts and individual phonemes as the properties of these words.

Nondeclarative memory includes:

• Habits, which are well-rehearsed, almost unconscious, behavioral routines
• Procedural memory (PM), which facilitates learning of motor tasks and procedures (playing golf) and developing cognitive skills (solving a Sudoku puzzle) by repeatedly performing them 
• Priming, which is an implicit or unconscious enhanced ability to detect, identify, or respond to a stimulus after recent exposure to it or to a related stimulus 
• Conditioned responses, which are "behaviors that are automatically produced in response to a particular set of stimuli" (Bayles & Tomoeda, 2007, p. 45). Culturally familiar items (food, music) evoke positive affect in dementia patients because of a lifetime of conditioned responses.

Memory Systems in Dementia

The specific neuropathology of Alzheimer's disease (AD) results in significant working and episodic memory impairments, but relatively spared nondeclarative memory. By contrast, patients with dementia due to Huntington's disease or Parkinson's disease have severe procedural memory and motor-skill-learning impairments, but less-impaired EM.

AD is the diagnosis for two-thirds of all individuals with dementia (Katzman & Bick, 2000). According to the Alzheimer's Association (2006), nearly 4.5 million Americans have AD and an estimated 11.3 to 16 million Americans will have AD by 2050. Deficits in EM appear first and are severe in AD, and are thus critical to identifying AD early. Semantic memory also is affected as AD progresses; however, there is compelling evidence that SM is better preserved than previously thought, especially in early AD. Available evidence on NDM function suggests that it is relatively spared until later stages of AD (Table 2 [PDF]).

In his book Searching for Memory, Dan Schacter describes these memory systems in a man with mild Alzheimer's-related dementia as revealed during a golf game. The man had been a highly skilled golfer for 30 years and retained a high level of skill (procedural memory). He excelled at driving, putting, choosing the appropriate clubs for his shots, and recalled the rules and strategies of golf. He demonstrated preserved semantic knowledge and correctly used golf-related vocabulary (birdies, doglegs, wedge, par, etc.). However, his EM deficits were severe; he could not recall the shot he had just hit and would re-attempt it. When complimented on a recent shot, he would be surprised and could not remember it. Shortly after the game, he could not recall details of the game he had just played, with whom he played, or that he had played that day.

In comparison, individuals with "semantic dementia" show severe, early-appearing SM deficits despite reasonably well-functioning EM. A recent client with semantic dementia could not explain meanings of common words (nail, pencil) or answer questions about physical attributes and functions of these objects. She also presented with a near-complete loss of knowledge of famous personalities (Marilyn Monroe, President John F. Kennedy) and historical facts. Yet, she easily and vividly recalled details of her son's recent wedding and her retirement celebration two years earlier.

Screening

SLPs can play a pivotal role in early identification and clinical diagnosis of dementia. When working with older adults, screening should briefly address five areas: 

• Review of medical history, emphasizing co-occurring medical conditions, current medications, recent history of surgeries and hospitalizations, physical/psychological trauma, behavioral problems, and symptoms of memory loss or confusion. Information from the medical chart, client/caregiver report, and observations may be used to obtain a Hachinski Ischemia Score (Hachinski et al., 1975), which helps rule out the possibility of dementia due to multiple infarctions. 
• Subjective report of memory problems from clients or caregivers can signal individuals who may be at a greater risk for developing dementia, even when objective screening measures indicate cognitive functions within normal limits (Wang et al., 2004). It is helpful to use a simple Likert scale to quantify responses to questions about self-reported memory impairments. 
• Hearing and vision function. Hearing loss (HL) affects one in three adults older than 60 and half of those older than 85 (National Institutes of Health, 2001). HL negatively affects communication, socialization, and quality of life in older adults and often mimics early signs of dementia. Identification and management of HL can "lessen the severity of a confusional state" (ASHA, 1997). Per current guidelines (ASHA, 2005), a hearing screening for clients with dementia should include otoscopy, pure-tone audiometry, an inspection of hearing aids (when applicable), and a cursory check of word-recognition abilities. Vision impairments also occur frequently in older adults; screening can be done using the Arizona Battery for Communication Disorders of Dementia (Bayles & Tomoeda, 1993), which includes tasks to assess scanning, object identification, visual perception, and literacy. 
• Cognitive function, especially of memory and executive function, is foundational to a dementia diagnosis. Therefore, documenting cognitive status is critical in any screening for dementia. The Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975) and the Clock Drawing Test (Sunderland, Hill, Mellow, Lawlor, Gundersheimer, Newhouse, & Grafman, 1989) are two widely used screening tools. 
• Depression, common in individuals with dementia, may contribute to poorer test performance (ASHA, 2005) and affect treatment outcomes. Therefore, a brief screening for depression using an instrument similar to the Geriatric Depression Scale Short Form (Sheikh and Yesavage, 1986) is warranted to determine the need for detailed assessment of affective status.

Comprehensive Assessment

When working with clients who have dementia, even if intervening for dysphagia, it is necessary to document cognitive status to justify a client's candidacy for skilled intervention and to determine the choice of the type of interventions. A client with moderate dementia and dysphagia may have sufficient alertness and residual procedural memory to benefit from a direct intervention to train safe swallowing strategies. A client with severe dementia and dysphagia, however, would not be a candidate for direct intervention. In this case, indirect interventions involving diet changes, caregiver training, and use of a functional maintenance plan are more justifiable.

Many standardized tests are available to clinicians and researchers working with clients who have dementia. The choice of a test or selected subtests of a test depends on the client's dementia severity and the type of information desired about the client. Table 3 [PDF] lists six client-centered questions and recommended tests/measures to answer each (these recommendations are not intended to be exhaustive but representative of available options).

Interventions

Direct interventions for patients with dementia should aim to facilitate new learning with a goal of maximizing daily functioning and quality of life. Facilitating learning in patients with dementia requires clinicians to understand spared and impaired memory systems and to apply principles that enhance encoding, storage, and retrieval of information.

Five evidence-based principles for enhancing learning in dementia (Mahendra, 2001) are described below. 

• Strengthen memory traces with repetition. Repetition is a simple, effective strategy to help patients with dementia learn new skills and information. Researchers have demonstrated that repetition enhances episodic recall in dementia and improves communication with memory-impaired clients. When learning information through repeated exposure, persons with dementia learn previously familiar (but forgotten) information sooner and retain it longer than novel information. It is important that clients have repeated opportunities to retrieve information or perform a procedure after it is presented or demonstrated by a clinician. 
• Engage the nondeclarative memory system. Active involvement promotes better learning. Clients with dementia can be actively involved by being asked to generate a response during a task rather than passively watching and listening. For example, when presented with pictures depicting a story, patients with dementia better remembered the story when they saw and described the pictures than when they only saw them. Similarly, a client being taught a procedure must repeatedly practice it to engage the relatively spared nondeclarative (learning-by-doing) system. Psychologists refer to this as the "generation effect"—better memory for self-generated responses than for observed responses. Generating a response requires effort and allows information to be processed at a deeper level, making the memory of the response more durable. 
• Reduce errors during learning. Minimizing errors during learning is essential because every time an error is repeated, this erroneous response is strengthened. Reducing errors is crucial for persons with AD who have severe EM deficits, because they cannot recall making the error and clinicians' feedback regarding errors. Additionally, clients with AD have inhibitory deficits (reflecting frontal lobe involvement) and cannot easily suppress irrelevant, erroneous responses. Given that the very act of committing an error reinforces it, errorless learning involves directing clients not to guess a response if they are unsure and designing interventions so that cues are provided at retrieval to constrain client responses. 
• Structure intervention so clients focus attention on a single task. Attention deficits appear early in AD and worsen as dementia progresses. For patients with dementia, divided-attention tasks (requiring simultaneously dividing one's attention between two or more tasks or processes) are harder than those requiring focus on a single task. Because clients with AD have divided-attention deficits and working memory limitations, engaging in multiple tasks at the same time overwhelms them. 
• Use salient cues to aid successful retrieval. Free recall is inherently harder than cued recall or forced-choice recognition tests. Cues enhance recall of information because they trigger a more focused search of episodic memory. For example, clients with dementia often struggle to freely recall a short story after a 10- or 15-minute delay; the cues in multiple-choice questions about the same story help them access information they cannot recall freely. Cues also help to decrease the frequency of undesirable behaviors and to increase the frequency of desirable behaviors (for example, a compensatory strategy). Bourgeois and colleagues (1997) effectively trained caregivers to use written cues to decrease repetitive verbalizations of their spouses with AD. Caregivers used written cues consisting of answers to their spouses' repetitive questions. When their spouses asked a recurring question, caregivers directed their spouses to read the cue cards. Further, to capitalize on spared capacity for conditioning of these spouses with AD, caregivers praised their spouses verbally when the cues were correctly used. These caregivers successfully reduced their spouses' repetitive verbalizations and maintained use of these cues over a six-month follow-up period. Similarly, Brush & Camp (1998) used a written cue card at mealtimes (during spaced retrieval training) to prompt a client with dementia and dysphagia to alternate sips of liquid with bites of food.

Human memory systems are multi-dimensional and incredibly complex. Although dementia is a progressive condition, it does not uniformly impact all memory systems. This finding is very exciting for clinicians who can use preserved memory abilities to justify the candidacy of dementia patients for comprehensive assessment and skilled interventions.

The very endeavor of speech-language treatment is to facilitate re-learning and new learning of information and skills. Therefore, SLPs must make every effort to apply current knowledge of memory systems to understand, document, and take advantage of the restorative potential of dementia patients to better design interventions.  

Preparation of this article was supported by a research grant awarded to the first author by the Alzheimer's Association.