CHAPTER 2 Consider the two language samples presented in Box 2-1. Both were produced by 75-year-old men who had completed 4 years of college education. Neither of the men had a history of neurological disease, diabetes, ischemic heart disease, significant hearing loss, or other major medical conditions. Speaker A is fluent and articulate. He expresses himself clearly with little repetition or redundancy; he uses a range of different grammatical structures and lexical items and few fillers. Speaker B is struggling to express himself; his speech is fragmented and marked by many repetitions and fillers. When he does manage to produce a complete sentence, it is short and grammatically simple. This chapter will consider a variety of explanations for the marked differences in the fluency, grammatical complexity, and linguistic content of Speakers A and B. Box 2-1 Language Samples from Speakers with Distinct Working Memory Capacities One difference between Speaker A and Speaker B is their working memory capacity. On conventional tests of working memory span, Speaker A scores very well, with span scores typically in the same range as those observed in young adults. His Forward Digit Span score was 7.6; his Backward Digit Span was 6.3; and he attained a score of 4.5 on a Reading Span test. In contrast, Speaker B has a more limited working memory capacity, with a Forward Digit Span score of 5.4; a Backward Digit Span of 3.0; and a Reading Span of 2.0. (These tests are more fully described in Appendix 2-1.) Working memory is essential to many everyday tasks that involve the retention of information; working memory has two functions: the short-term retention of information and the manipulation of information. The prevailing model of working memory, as proposed by Baddeley (1986) and Baddeley and Hitch (1974), involves three components: two temporary storage mechanisms that buffer visual information (e.g., the visual scratchpad) and auditory information (e.g., the phonological loop) and a central executive processor. A fourth component, an episodic buffer linked to long-term memory, was added by Baddeley (2000). Cowan (1995, 2001), McElree (2001), and Oberauer and Kliegl (2006), among others, have proposed mixture models linking attention and working memory. A chief characteristic of these multicomponent systems is that the system has limited capacity – to temporarily store information or to divide attention among processing tasks. Each component is also assumed to have unique characteristics: the auditory buffer is speech-based, while the visual buffer is spatially defined. Executive function itself has been typically defined very broadly, as “those capacities that enable a person to engage successfully in independent, purposive, self-serving behavior” (Lezak, Howieson, Loring, et al., 2004, p. 35), or as “a multidimensional construct referring to a variety of loosely related higher-order cognitive processes including initiation, planning, hypothesis generation, cognitive flexibility, decision-making, regulation, judgment, feedback utilization, and self perception” (Spreen & Strauss, 1998, p. 171), and as a bundle of “general purpose control mechanisms that modulate the operation of various cognitive subprocesses” (Miyake, Friedman, Emerson, et al., 2000, p. 50). Executive component itself includes different functions such as attentional allocation and selection, inhibition, and information updating. One challenge to understanding the role of working memory in language and communication is the multiplicity of tests and assessments used to measure individual differences in working memory. Working memory is typically defined by tests of working memory span and by tests of executive function (Box 2-2). Executive function itself is measured by neuropsychological tests such as the Wisconsin Card Sorting Test, or specific tests of inhibition, time-sharing, updating, and switching. These tests are briefly described in Appendix 2-1. Many variants of each test have been developed. In addition, the speed of information processing in working memory also affects language and communication and a variety of approaches have been used to assess processing speed. There is considerable debate as to whether these tests assess separate but correlated executive functions or a unitary construct, and their relationship to general intelligence. For example, Engle, Tuholski, Laughlin, and Conway (1999) suggest that span tasks may be differentiated into simple span tests assessing short-term memory and complex span tests involving executive processes. A further issue is working memory can be subdivided into verbal and nonverbal (or visual/spatial) domains. See also the “users guide” developed by Conway, Kane, Bunting, et al. (2005) for a discussion of many methodological and procedural problems in measuring working memory capacity using counting, operational span, and reading span tests. There is no single measure that serves as the “gold standard” for the assessment of executive function. Salthouse, Atkinson, and Berish (2003), noting the complexity and breadth of notions of executive function, undertook an examination of the construct validity of executive function in a sample of 261 adults ranging in age from 18 to 84 years. Their approach was to examine convergent and discriminate validity among a set of neuropsychological and cognitive tasks typically associated with executive function, and also a set of psychometric tasks including measures of verbal ability, fluid intelligence, episodic memory, and perceptual speed. A series of structural equation analyses were then conducted to look at the relations among these sets of variables. Their results indicated that the various neuropsychological measures were not very highly related to one another and were fairly highly related to other variables, particularly fluid intelligence. They concluded that individual differences in measures of working memory may in fact reflect differences in much broader abilities, such as fluid intelligence. Miyake and colleagues (Friedman & Miyake, 2004; Miyake et al., 2000) have addressed similar questions but took a somewhat different approach and reached different conclusions. Miyake et al. (2000) reported a study addressing “the unity and diversity of executive functions” (p. 49) using confirmatory factor analysis and structural equation modeling. They found that a three-factor solution fit the data better than any of the one- or two-factor solutions, indicating that there are three separable dimensions of executive function. The authors conclude from this study that the three executive functions they measured (updating, shifting, inhibition) are “clearly distinguishable” and that each plays a different role in more complex executive function measures such as the Wisconsin Card Sorting Test and the Tower of Hanoi. The questions of how aging affects working memory have not been clearly answered by either the Salthouse study or the Miyake studies. As measured by simple and complex span tests, working memory increases in childhood (Dempster, 1980; Gathercole, 1999; Park, Smith, Lautenschlager, et al., 1996; Pickering, 2001) and declines in late adulthood. What drives this U-shaped increase, then decrease in working memory is the subject of considerable debate. Salthouse (1994, 1996) has argued for processing speed as the fundamental mechanism; Lindenberger and Baltes (1994; Baltes & Lindenberger, 1997) have argued for neural integrity as measured by sensory acuity and postural balance and gait as the critical factor; and Hasher and Zacks (1988) have argued for a breakdown in inhibitory functions. Inhibition is critical for blocking irrelevant information from entering working memory, deleting irrelevant information from working memory, and restraining prepotent responses. Under this hypothesis, older adults with poor inhibitory mechanisms may not only be more susceptible to distraction, but they may also be less able to switch rapidly from one task to another and they may rely on well-learned “stereotypes, heuristics, and schemas” (p. 123) (Yoon, May, & Hasher, 1998). Lustig, May, and Hasher (2001) have demonstrated that working memory span in older adults can vary dramatically as a function of test format, comparing the traditional format for testing memory span uses a sequence of trials in which set size increased from 2 to 3 to 4 to 5 items with a format designed to minimize interference in which set size decreased from 5 to 4 to 3 to 2 items. Whereas this manipulation did not affect span estimates for young adults, it did for older adults, implying that the traditional test measures not only working memory capacity but also inhibition. Another issue is whether cognitive abilities dedifferentiate with age, becoming more highly correlated (Cornelius, Willis, Nesselroade, & Baltes, 1983; Li, Lindenberger, Homnel, et al., 2004). Dedifferentiation is assumed to arise from the decline of a basic, fundamental mechanism, such as processing speed, whereas differentiation is assumed to arise from the development or breakdown of process-specific mechanisms. Rabbitt and Lowe (2000; Rabbitt, 1993) have suggested that aging leads to increasing individual differences, reflecting different rates and trajectories of change of underlying processes and/or neural structures. To investigate these questions, Hull, Martin, Beier, et al. (2008) used an approach similar to that used by Miyake et al. (2000) and administered a battery of tests of shifting, updating, and inhibition to a panel of middle-aged and older adults along with two criterion tests of executive function, the Wisconsin Card Sorting Test and the Tower of Hanoi test and tests of verbal and nonverbal knowledge. Their analysis suggested two underlying factors: shifting and updating with minimal overlap between these two factors. Performance on the two criterion tests was best predicted by updating, the ability to maintain information and track rule changes in working memory. Shifting, the ability to activate alternative rules, did not contribute to performance on the Card Sorting and Tower of Hanoi tests. Somewhat surprisingly, Hull et al. found no evidence for a third inhibition factor, perhaps due to a lack of measurement sensitivity for the Stroop and antisaccade tests used to assess inhibition. They also note that aging appears to affect the relative contributions of underlying factors; in the Miyake et al. (2000) study, shifting was the primary predictor of performance on the Card Sorting test whereas Hull et al. found that updating was the best predictor. Thus, as aging affects different components of working memory, the relative balance among preserved components may be altered. A decline in working memory capacity may lead increased reliance on efficiency; hence executive function in younger adults may be more dependent on the capacity to store multiple representations in working memory whereas executive function in older adults may be more dependent on the efficiency at which information can added to or deleted from a (reduced) working memory. A similar conclusion was reached by McDowd et al. (2011) in a recent study that compared how young and older adults’ performance on a variety of verbal fluency tests covaried with other measures of cognition including measures of processing speed, inhibition, working memory capacity, and verbal ability. Letter fluency (e.g., words beginning with “M”), semantic fluency (e.g., “colors” or “fluids”), and action fluency (e.g., “ways you can talk”) were tested; processing speed was assessed by performance on the digit symbol and letter comparison tests (see later); working memory capacity was measured by forward and backward digit span and by reading span; inhibition was determined by performance on the Wisconsin Card Sorting Test and on the Stroop and Trail-Making Tests; and verbal ability was measured by performance on the Boston Naming (Kaplan, Goodglass, & Weintraub, 1983) vocabulary test. The group differences were very similar across fluency measures and types: in general, young adults produced more correct responses, fewest perseverations, and fewest intrusions and the older adults produced fewer correct responses, more perseverations, and more intrusions. To examine how individual differences in processing speed, verbal ability, working memory, and inhibition affected performance on the verbal fluency tests, a series of regression models was evaluated separately for the young and older adults. For young adults, these models were nonsignificant, perhaps reflecting the restricted range of young adults’ performance on these tests but also supporting the differentiation of cognitive abilities into separable components. For older adults, processing speed and inhibition were the best predictors of performance on the fluency tests, suggesting that the speed of information retrieval from semantic memory as well as the ability to select and focus retrieval operations are key determinates of verbal fluency. Vocabulary size and working memory capacity do not appear to affect older adults’ ability to retrieve letter, category, and action exemplars whereas processing speed and inhibition do, perhaps because speed and efficiency become a more critical determinates of verbal fluency performance as aging leads to declines in working memory capacity as well as increases in vocabulary. Working memory and executive function are believed to be subserved by the prefrontal cortex (Raz, 2005) and the nigrostriatal dopamine neurotransmitter system (Arnsten, Cai, Steere, & Goldman-Rakic, 1995; Volkow, Wang, Fowler, et al., 1998). Both are affected by aging and, in turn, affect performance on working memory and executive function tasks. Although there is an overall reduction in brain volume with advancing age, this loss is accelerated in the prefrontal cortex (Dennis & Cabeza, 2008; Raz, 2005; Raz, Gunning, Head, et al., 1997; Raz, Lindenberger, Rodrigue, et al., 2005; Salat, Kaye, & Janowsky, 1999), arising from neuronal shrinkage and declines in synaptic density (Huttenlocher & Debholkar, 1997; Peters, Morrison, Rosene, & Hyman, 1998). This loss of brain volume results in reduced prefrontal activation (Grady, McIntosh, & Craik, 2005; Grady, McIntosh, Rajah, et al., 1999). Figure 2-1 compares regional brain volume changes in healthy adults. Although many neurotransmitter systems are affected by aging, the most dramatic changes appear in the dopamine system (Bäckman & Farde, 2005; de Keyser, Herregodts, Ebinger, et al., 1990; Suhara, Fukuda, & Inoue, 1991; Volkow et al., 1998). Age-related declines in the dopamine system, in turn, result in reduced input to the frontal cortex, reflecting the functional interconnectedness of a frontalstriate circuit (Volkow et al., 2000). These dopaminergic pathways are illustrated in Figure 2-2. Cabeza (2002) has proposed that frontal activity is less strongly lateralized in older adults than in young adults, implying that older adults compensate for neurocognitive deficits by recruiting both hemispheres to perform tasks that require only a single hemisphere in young adults. This pattern of age-related asymmetry reductions appears during tests of paired associate learning (Cabeza, McIntosh, Tulving, et al., 1997), word stem recall (Bäckman, Almkvist, Andersson, et al., 1997), and word recognition (Madden, Langley, Denny, et al., 2002), as well as on verbal working memory tests (Reuter-Lorenz, Jonides, Smith, et al., 2000). In addition to changing the brain’s structure and organization, aging may affect the neurochemical basis of cognition by altering or modulating signal transmission between and among neurons (Li, 2005; Li & Silkström, 2002). As a result, stimulus-response relations may be altered, reducing sensitivity to stimuli, increasing the temporal variability of responses, and increasing “noise” or random activations (Li, Lindenberger, & Fransch, 2000). One further consequence may be that events and stimuli are encoded less distinctively, resulting in a blurring of episodic memories, an increasing in the variability of performance, and the dedifferentiation of cognitive abilities as they become more correlated (Li et al., 2005). There is widespread agreement that working memory is critical to a wide range of cognitive abilities that affect older adults’ language and communication. Support for the hypothesis that working memory limitations constrain older adults’ production and comprehension of language are largely correlational. A variety of observations support this hypothesis; for example, performance on the reading and listening span tests of Daneman and Carpenter (1980) has been shown to be related to performance on reading and listening comprehension, learning to read, reading ability, arithmetic ability, and reasoning ability (Daneman & Blennerhassett, 1984; Daneman & Green, 1986; Daneman & Tardif, 1987; Hitch et al, 2001; Leather & Henry, 1994). Daneman and Merikle (1996) reviewed 77 studies involving 6,179 participants, confirming the link between reading/listening span measures and language comprehension, reporting correlations of .41 and .52 with global and specific tests of comprehension. Older adults have typically been found to have smaller working memory spans than young adults and such span measures have been found to correlate with measures of language processing (Borella, Carretti, & De Beni, 2008; Norman, Kemper, Kynette, et al., 1991; Stine, Wingfield, & Myers, 1990; Tun, Wingfield, & Stine, 1991). One approach has been to examine the relationship between measures of language production, obtained from elicited language samples, and measures of working memory, obtained from span or other tests. Language sample analysis relies on a variety of metrics to evaluate language including measures of fluency, grammatical complexity, and content. Typical metrics are summarized in Box 2-3 and further described in Appendix 2-2; the application of the DLevel and PDensity metrics is illustrated in Appendix 2-3. A variety of specialized software is available to assist with language sample analysis including the Systematic Analysis of Language Transcripts (SALT) software developed by Chapman and Miller (1984) and the Computerized Propositional Idea Density Rater (CPIDR) (Brown, Snodgrass, Kemper, et al., 2008). In addition, the on-line calculator Coh-metrix (Graesser, McNamara, Louwerse, & Cai, 2004) may be used to obtain additional measures; although originally developed to assess the coherence of written documents, it may be used to conduct analyses of elicited language samples. Table 2-1 illustrates the application of these metrics by comparing the two language samples from Box 2-1, one elicited from an older adult with excellent working memory and one elicited from an older adult with poor working memory, as well as a third language sample discussed later. Table 2-1 Application of Language Sample Metrics to 3 Examples Cheung and Kemper (1992) used structural modeling to investigate interrelationships among many language sample metrics as well as measures of working memory capacity and verbal ability using language samples elicited from young and older adults. Cheung and Kemper showed that age-related declines in working memory were highly correlated with age-related declines in grammatical complexity assessed by metrics that are sensitive to the length of grammatical constituents, how many clauses are embedded within a sentence, and how those clauses are embedded. Kemper and Sumner (2001) extended this approach to investigate the relationship between language sample measures and traditional measures of verbal ability, working memory, and verbal fluency. They reported that grammatical complexity was correlated with span measures of working memory. In contrast, propositional content was correlated with measures of verbal fluency and reading rate, suggesting that processing speed and efficiency limit how information can be conveyed linguistically. Verbal ability, assessed by performance on vocabulary tests, constituted a third factor unrelated to the language sample measures of grammatical complexity or propositional content. These correlational analyses suggest that working memory imposes a ceiling on how many sentence relations can be formulated at one time. Each embedded or subordinate clause increases the burden on working memory by imposing additional requirements for subject-verb agreement, pronominal choice, the linear ordering of adjectives, and the application of other grammatical rules. If older adults’ language production is functionally limited, it should be evident in how young and older adults respond in controlled production experiments in which participants are given words or sentence fragments and asked to compose a sentence. In a series of studies, Kemper, Herman, and Lian (2003a) and Kemper, Herman, and Liu (2004) varied the number of nouns and the types of verbs given to the participants and scored the length, grammatical complexity, and propositional or informational content of each sentence produced and the time taken to respond. Older adults’ responses were similar to those of younger adults when given 2 or 3 words. When given 4 words, the older adults were slower to respond, made more errors, and their responses were shorter, less complex and less informative than the younger adults’ responses. When different types of verbs were provided, young and older adults responded similarly with simple intransitive (smiled) and transitive (replaced) verbs but older adults encountered problems using verbs like expected that preferentially are used with embedded clauses, e.g., . . . expected the package to be delivered. Older adults responded very slowly yet produced shorter, grammatically simpler, and propositionally less informative sentences. Other researchers have shown that working memory limitations affect older adults’ language processing using tests of text comprehension and recall. Kwong See and Ryan (1996) examined whether individual differences in text processing are attributable to working memory capacity, processing speed, or the breakdown of inhibitory processes. Working memory capacity was estimated by backward digit span, processing speed by color naming speed, and inhibition by performance on the Stroop task. Their analysis suggested that older adults’ text processing difficulties can be attributed to slower processing and less efficient inhibition, rather than to working memory limitations. Van der Linden et al. (1999) also sought to distinguish the effects of working memory limitations from those due to reductions of processing speed or a breakdown of inhibitory processes by examining performance on a wide range of language tasks using structural equation modeling. Young and older adults were tested on their ability to understand texts and recall sentences and words. They were also given a large battery of tests designed to measure processing speed, working memory capacity, and the ability to inhibit distracting thoughts. The analysis indicated that these three general factors (speed, working memory, inhibition) did account for age-differences in performance on the language processing tasks. Further, the analysis indicated that “age-related differences in language, memory and comprehension were explained by a reduction of the capacity of working memory, which was itself influenced by reduction of speed, [and] increasing sensitivity to interference . . . ” (p. 48). A limitation of these studies is age-related changes to language processing are inferred from performance on recall measures, answers to comprehension questions, or global measures of reading speed. A more specific set of hypothesis about the nature of age-related changes to language processing have been examined in a series of studies using more “direct” methods to examine the role of working memory in syntactic processing. Just and his colleagues (Just & Carpenter, 1992; Just & Varma, 2002; King & Just, 1991; MacDonald, Just, & Carpenter, 1992) have claimed that working memory capacity constrains the interpretation of temporary syntactic ambiguities, limiting the ability of older or low span readers to make and sustain multiple interpretations of ambiguous phrases. According to the Just and Carpenter (1992) capacity-constrained (CC) theory (see also the 3CAPS model of Just & Varma, 2002), older or low span readers should have difficulty processing temporary syntactic ambiguities and should exhibit garden-path effects, initially misinterpreting reduced relative clause constructions as main verbs only to reinterpret the constructions once disambiguating information is encountered. Young or high span readers should be able to avoid garden-path effects, by constructing multiple syntactic interpretations of the ambiguous phrases and retaining these interpretations until disambiguating information is encountered. This hypothesis has been carefully examined by Caplan and Waters (1999) who have considered a number of lines of evidence from studies of young and older adults as well as individuals with aphasia and dementia. They distinguish between immediate, interpretive syntactic processing and post-interpretative semantic and pragmatic processing. Caplan and Waters argue that there is little evidence to support the hypothesis that working memory limitations affect immediate syntactic processes; rather, they conclude that working memory limitations affect postinterpretative processes involved in retaining information in memory in order to recall it or use it (e.g., to answer questions or match sentences against pictures). In a variety of studies comparing adults stratified into groups based on measures of working memory, Caplan and Waters (1999) note that effects of syntactic complexity do not differentially affect high versus low span readers or listeners. And they report that secondary tasks that impose additional processing demands on working memory do not differentially affect the processing of complex sentences. Caplan and Waters consider aphasic patients such as B. O. who had a digit span of only 2 or 3 digits but who was able to perform as well as normal healthy older adults on a wide range of tasks with complex sentences. They also note that patients with Alzheimer’s dementia, who also show severely limited working memory capacity, are able to make speeded acceptability judgments of complex sentences as accurately as nondemented controls. Waters and Caplan (1996a, 1996b, 1997, 2001) have directly examined the hypothesis that working memory limitations affect older adults’ ability to process complex sentences. These studies have used the auditory moving windows paradigm. This technique allows the listener to start and stop the presentation of sentence and permits the analysis of phrase-by-phrase listening times, analogous to visual moving windows paradigms, which permit the analysis of word-by-word or phrase-by-phrase reading times. The studies by Caplan and Waters typically examine the processing of subject and object relative clause constructions, such as those that follow:
The effects of aging on language and communication
Working memory, aging, and language processing
Concept of working memory
Measuring working memory
Aging and working memory
Aging and the neurological basis of working memory
Working memory constraints on language processing
Speaker A
Speaker B
Speaker C
MLU
10.00
4.70
4.86
MCU
1.35
0.48
0.83
DLevel
2.50
1.13
0.21
PDensity
5.82
2.02
5.13
Speech rate, wpm
124 wpm
96 wpm
84 wpm
TTR
.62
.87
.59
Fragments, %
20
71
66
A ceiling on language production
Syntactic processing limitations
Related posts:
The cognition of language and communication
Management of communication deficits associated with memory disorders
The role of memory in language and communication
Management of acquired language disorders associated with attentional impairment
Language and communication disorders associated with attentional deficits
Attention: architecture and process
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
