| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
Department of Psychology, Brandeis University, Waltham, Massachusetts.
Address correspondence to Lisa M. Soederberg Miller, Brandeis University, Department of Psychology, Waltham, MA 02454-9110. E-mail: lmiller{at}brandeis.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionAppendixReferences |
|---|
ALTHOUGH age-related declines in speed of processing and working memory capacity are commonly reported in the literature (e.g., Salthouse & Babcock, 1991
; cf. Salthouse, 1991), knowledge stores remain constant or increase in late life (Schaie, 1990), leading some to argue that the accumulation and use of knowledge become increasingly important in late life (cf. Labouvie-Vief, 1985). Text comprehension is a particularly important facet of cognition as it is required for numerous daily activities. Although research clearly indicates that knowledge enhances memory for text, less is known about how knowledge affects encoding processes and whether there are age differences in these processes. In the present research, I investigate age differences in the effects of knowledge on encoding and recall of text.
Effects of Knowledge on Memory for Text
A good deal of research supports the notion that knowledge enhances the comprehension and recall of text (Chiesi, Spilich & Voss; 1979; Kintsch, 1988; Spilich, Vesonder, Chiesi, & Voss, 1979; Voss, Vesonder, & Spilich, 1980). The more one knows about a particular topic, the more readily he or she can draw inferences and identify concepts and integrate them to form a meaningful representation of the text. This well-formed representation then can be used to enrich memory. Several researchers have argued that retrieval structures are the mechanisms responsible for these benefits (Ericsson & Kintsch, 1995). Retrieval structures are created during encoding and provide individuals with a stable and rich representation of the text that resides in long-term working memory. Material in long-term working memory has the storage capacity of long-term memory and the accessibility associated with working memory. Thus, retrieval structures provide experts with a powerful memory aid.
However, it is less clear whether there are age differences in the effects of knowledge on memory performance. Because of the advantages associated with knowledge, several researchers have asked whether additional years of experience can offset age-related declines in cognitive processing and, in turn, reduce age differences in performance (e.g., Salthouse, 1987). Unfortunately, evidence supporting differential benefits of experience and knowledge among older adults is mixed. For example, some research has shown that experience differentially benefits older adults (e.g., Clancy & Hoyer, 1994; Morrow, Leirer, Fitzsimmons, & Altieri, 1994) whereas other research indicates that it does not (e.g., Morrow, Menard, Stine-Morrow, Teller, & Bryant, 2001; Salthouse, Babcock, Skovronek, Mitchell, & Palmon, 1990). The question of whether knowledge and experience can compensate for age-related declines in processing has also been posed within the literature on text processing (Dixon & Bäckman, 1993). The data here are also mixed, with some research showing that the elderly are differentially aided by knowledge and experience (Hess, 1990; Hess, Donley, & Vandermaas, 1989; Hultsch & Dixon, 1983) and other research showing age constancy (Arbuckle, Vanderleck, Harsany, & Lapidus, 1990; Hambrick & Engle, 2002; Morrow, Leirer, & Altieri, 1992). Thus, the data indicate that knowledge derived from experience improves recall performance for older adults at least as well as it does for younger adults.
Effects of Knowledge on Encoding of Text
Although research demonstrates that knowledge improves memory performance, less is known about how knowledge affects encoding processes. Some research indicates that knowledge reduces the time required to process text (Miller & Stine-Morrow, 1998; Sharkey & Sharkey, 1987). Sharkey and Sharkey (1987) demonstrated that readers spend less time processing sentence-final target words when they are able to draw upon script knowledge presented earlier in the text. The location of this facilitation is significant because readers pause at the ends of major constituents to organize and integrate, that is "wrap-up," the previously read text (Aaronson & Scarborough, 1976; Just & Carpenter, 1980). Wrap-up processing consists of several reading processes including conceptual organization and integration and ambiguity resolution (Aaronson & Scarborough, 1976; Just & Carpenter, 1980). Thus, Sharkey and Sharkey's finding suggests that script knowledge reduces the cognitive demands associated with wrap-up processing.
Miller and Stine-Morrow (1998) also investigated the effects of knowledge on reading times. We randomly assigned knowledge by using a Bransford and Johnson-style (1972) manipulation in which half the participants received passage titles that transformed otherwise vague and ambiguous passages into easily recognized schemas. The passage titles therefore provided useful knowledge for understanding the texts. Miller and Stine-Morrow (1998) found that high-performing young and elderly readers spent less time performing wrap-up when readers were given this knowledge. Interestingly, this effect was more pronounced among the older readers, suggesting that older adults were differentially facilitated by knowledge during encoding. Thus, these data suggest that script and schematic knowledge reduce the demands associated with text processing and that this is particularly true for older adults.
However, not all research has shown that knowledge facilitates processing. Research investigating the effects of domain knowledge on reading suggests that knowledge application can be effortful (Britton & Tesser, 1982). Britton and Tesser (1982) found that readers with background knowledge were slower to respond to a secondary task than were readers who did not possess relevant background knowledge. Presumably, this increased time reflects the demands made by the primary task of reading. The researchers developed the "prior knowledge hypothesis" to account for this finding and argue that this increased effort occurs because knowledge must be "unpacked." Consistent with this hypothesis, readers who knew more about cooking allocated more time to wrap-up processing of cooking texts than did those who knew less about cooking (Miller, 2001). These data suggest that knowledge engenders the allocation of cognitive resources to domain-relevant texts. Knowledge use may involve greater allocation of resources because the more one knows about a particular domain, the more connections there are to be made between prior knowledge and concepts presented in the text. The allocation of resources to text, as conceptualized by Graesser (1981), refers to a methodological approach used to determine how time is distributed to various text processes. Although the nature of these resources is not entirely clear, there is some evidence to suggest that they may not be related to working memory capacity as measured by tasks such as a loaded span task (Smiler, Gagne, & Stine-Morrow, in press; cf. Caplan & Waters, 1999; Waters & Caplan, 2002).
Focus of the Present Study
The focus of the present research was to further investigate encoding processes associated with the use of knowledge and to determine whether there are age differences in these processes. In particular, this research investigated age and knowledge differences in the amount of time readers allocate to conceptual organization and integration at wrap-up (Miller, 2001; Miller & Stine-Morrow, 1998; Sharkey & Sharkey, 1987). Some have argued that the use of knowledge may be particularly effortful for older adults (Arbuckle et al., 1990). Although some data suggest that knowledge can engender the allocation of resources among older adults (Miller, 2001), age comparisons of the effects of knowledge during encoding have been limited. To the extent that older adults differentially rely on knowledge to support text processing (Dixon & Bäckman, 1993), older adults should increase effort more than younger adults when applying knowledge. Specifically, knowledgeable older adults, relative to their younger counterparts, would show larger increases in resource allocation to wrap-up while reading domain-related texts. This hypothesis is consistent with life-span theorists who argue that with age comes a more contextual processing style that draws more heavily on past experiences and knowledge (e.g., Labouvie-Vief, 1985). A secondary goal was to determine whether knowledge could reduce age differences in memory for domain-related texts.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
On the basis of a cooking test (see below), participants within each age group were assigned to either a high- or low-knowledge group (cf. Chiesi et al., 1979). Because a one-way analysis of variance (ANOVA) on these scores showed that there were no age differences in cooking knowledge, a median split was conducted across age groups. The sample consisted of 15 low-knowledge (LK) and 16 high-knowledge (HK) younger adults and 31 LK and 24 HK older adults. The average age of the HK younger participants was somewhat higher than that of the LK younger participants, t(29) = 3.11, p <.01. Because correlations within this age group showed that age was not correlated to either central dependent variable (resource allocation or recall), this asymmetry should not pose a problem. The mean ages of the two older groups were comparable, t < 1.
To determine whether age and knowledge groups differed in ability, I performed a series of Age x Knowledge ANOVAs on measures of vocabulary, working memory, and short-term memory. I assessed vocabulary with the Shipley Institute of Living Scale (Shipley, 1940), which consists of 40 multiple-choice items. As can be seen in Table 1, older adults had higher vocabulary scores than did younger adults, F(1,82) = 18.74, p <.001. I assessed working memory capacity using a loaded reading span task based on Daneman and Carpenter (1980; cf. Stine & Hindman, 1994). This task required participants to read a successively longer list of unrelated sentences, respond true or false after each, and then recall the last word from each sentence in the order in which it was presented. In past research (Salthouse & Babcock, 1991), a comparable task was shown to have a reliability (split-half correlation) of.86. Younger adults scored higher on the loaded reading span task than did older adults, F(1,82) = 4.12, p <.05. I assessed short-term memory span with the forward and backward digit spans of the Wechsler Adult Intelligence Scale (WAIS; Wechsler, 1981). For these tasks, participants listened to and then repeated progressively longer strings of digits in the same (forward digit span) or reverse (backward digit span) order; the subscale combining these two tasks has been shown to have a test–retest reliability of.83 (Wechsler, 1981). Younger and older adults did not differ in their forward or backward digit spans, F < 1, for both.
|
Materials
Cooking knowledge test
I used a 20-item multiple-choice cooking test (Miller, 2001; Soederberg, 1997) to divide the sample into high- and low-knowledge groups. The test was developed from a battery of 100 items (Cronbach's 
=.93) and shortened to 20 items (Cronbach's =.88) during the validation process (cf. Soederberg, 1997). The items originated from cooking textbooks and cookbooks and represented a wide range of cooking knowledge from properties of sauces to principles of baking.
Passages
Reading materials consisted of four cooking passages and four control passages about general biology topics. Passages were between 200 and 204 words in length, contained between 83 and 93 propositions (Turner & Greene, 1978), and contained between 24 and 39 new concepts. Two of the control passages came from American Scientist articles, and two came from Stine-Morrow, Loveless, and Soederberg (1996), originally based on science and nature articles. Two cooking passages came from cooking textbooks, as described in Miller (2001), and I wrote two for use in this study. To ensure that both cooking and control passages were comparable in terms of perceived level of difficulty, a separate sample of 10 younger and 7 older participants (none of whom were selected for their cooking experience) rated each passage on an 8-point scale (1 = not at all difficult; 8 = extremely difficult) representing how difficult the passages would be to remember. Paired sample t tests on the mean within each condition showed that the cooking and control passages did not differ on perceived difficulty (M = 4.71 for cooking; M = 4.72 for control), t < 1, and this was true for younger, t(9) = 1.05, and older, t(6) = 1.12, adults.
Procedure
Passages within passage-type condition were blocked together in groups of two, and order of block was determined randomly. Participants read passages word-by-word using the moving-window technique (Just, Carpenter, & Woolley, 1982), such that each word appeared on the screen by itself. Successive button presses caused the location of the words to move from left to right across the screen and downward as with free reading. The beginning of a passage was marked with a plus sign to signal where the first word would appear, and the end of the passage was marked with "the end" and a cue to recall the passage. Participants were instructed to read at a comfortable reading rate that would enable them to recall the passage after reading it. Participants recalled passages out loud into a tape recorder for later transcription. A practice passage was given to familiarize them with the procedure. After reading four of the eight passages, participants were given a break from the computer, followed by the forward and backward digit spans and the vocabulary task. The sentence span task was administered at the end of the reading task, followed by a measure of passage interest that I included because past research has shown that interest in a particular area is related to knowledge in that area (Ackerman & Rolfhus, 1999). Participants rated their interest in each of the passages on a 5-point rating scale (1 = uninteresting; 5 = very interesting) in response to passage titles that were listed in alphabetical order on a sheet of paper. Because preliminary analyses indicated that associations between interest in the cooking passages and knowledge-related variables were unrelated, I dropped this variable from subsequent analyses. At the end of the session, participants completed the cooking knowledge test.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
).
Individual regressions
I assessed resource allocation by measuring reading times and estimating time spent on wrap-up processing by using regression analyses that controlled for the effects of nuisance variables (cf. Stine-Morrow et al., 1996). I performed two regressions for each participant (cf. Lorch & Myers, 1990) that predicted reading times from the characteristics of the cooking and control texts. For sentence wrap-up, words were coded in terms of whether or not they fell at a sentence boundary. A word was dummy coded with a 1 if it occurred at the end of a sentence and 0 if it did not. To obtain an estimate of wrap-up processing that was independent of other processes, several other text characteristics were coded so their effects could be partialled out of the regression equations. To control for orthographic coding, each word was coded in terms of the number of letters it contained. Words were dummy coded 0 or 1 for whether they occurred at the beginning of the sentence (sentence initiation), whether they represented a new concept in the passages (immediate processing), whether they occurred at the beginning of a new line of text (time to realign sight to the beginning of a new line on the computer screen), as well as the word's serial position within the passage (increases in reading rate that occur, for example, as a result of becoming familiar with contents of the passage). I removed from reading time analyses individuals who had regression coefficients associated with wrap-up that were greater than 4 SDs above the mean (within age and knowledge groups). This resulted in the loss of one older and one younger participant. Consistent with past research (Stine-Morrow et al., 2001), a reliability estimate of wrap-up processing calculated across the two passage conditions indicated that resource allocation to wrap-up processing was a reliable measure, =.80.
Wrap-up processing
I conducted an Age (young, old) x Knowledge (HK, LK) x Passage Type (cooking, control texts) repeated measures ANOVA on the regression coefficients representing sentence wrap-up for both passage types. Consistent with the prediction that knowledge would require an increase in resource allocation, there was a significant interaction between Knowledge and Passage Type, F(1,81) = 5.74, p <.05. As can be seen in Figure 1, HK readers increased resource allocation to domain-related texts but not to control texts. Follow-up tests showed that the difference in resource allocation to the cooking and control passages was nonsignificant for the LK readers, t(44) = 1.37, but was significant for the HK readers, t(39) = 2.32, p <.05. Neither the main effects of Knowledge, Age, and Passage Type, nor the two-way Age interactions reached significance, F < 1, for all. A nonsignificant three-way interaction, F < 1, failed to support the hypothesis that knowledge is particularly effortful for older adults.
|
in which participants received 0 points if they were unable to recall the proposition, 1 point if they recalled half of the gist of the proposition, and 2 points if they were able to recall the gist of proposition in its entirety. A second rater scored 20% of the protocols to ascertain the reliability of the scoring method. The correlation was.95, indicating a high degree of agreement between the two raters. The proportion of propositions recalled by each individual for the four passages within each condition showed good reliability for the cooking ( =.90) and control ( =.89) passages. I computed a summary score for each condition by taking the average proportion of propositions correctly recalled across the four passages. I conducted an Age x Knowledge x Passage Type repeated measures ANOVA on recall scores of the cooking and control passages. There was a main effect of Age, F(1,81) = 12.48, p <.01, such that younger adults (M =.47, SD =.10) recalled more than did older adults (M =.38, SD =.12). A main effect of Passage Type, F(1,81) = 100.13, p <.001, indicated that, in general, recall scores of the cooking passages (M =.45, SD =.17) were higher than those of the control passages (M =.37, SD =.12). This finding was surprising given that the passages had been equated on a number of variables, including length, propositional density, and perceived difficulty. The Passage Type x Age interaction was also significant, F(1,81) = 11.45, p <.01, indicating that the tendency to recall the cooking passages better than the control passages was more pronounced among the young than among the old (cf. Table 1).
The main effect of Knowledge on recall was significant, F(1,81) = 6.21, p <.05, showing that, in general, HK readers (M =.45, SD =.12) had higher levels of recall relative to LK readers (M =.38, SD =.10). It could be that the higher vocabularies of the HK group affected recall of the information presented in both types of passages. In fact, after I statistically controlled for vocabulary, this effect became nonsignificant, F(1,80) = 2.73.
It is important to note the main effect of Knowledge was qualified by a significant Knowledge x Passage Type interaction, F(1,81) = 4.09, p <.05, depicted in Figure 2. This interaction shows that, although both groups recalled more of the cooking passages, this increase was greater for the HK readers, t(39) = 7.77, p <.001, than for LK readers, t(44) = 4.76, p <.001. The Age x Knowledge x Passage Type interaction was nonsignificant, F < 1, indicating that the benefits of knowledge on recall were similar across age groups.
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
How Knowledge Affects Encoding Processes
A large body of research demonstrates that knowledge enables faster processing (cf. Ericsson & Smith, 1991); however, data from the present study indicate that the use of knowledge can involve an increase in processing time. These data are consistent with Britton and Tesser's (1982) prior knowledge hypothesis, past research on inference generation (Fincher-Kiefer, 1992), and past research on resource allocation to text among older adults (Miller, 2001). When knowledgeable individuals read texts for which they had relevant knowledge, they spent more time performing wrap-up processing relative to individuals without relevant background knowledge. This extra time may have been used to establish connections between the text and knowledgeable readers' more extensive knowledge base. For example, when cooks read about processes underlying sauces, they may already have an intuitive understanding of how sauces are prepared and how they can be ruined. As they read, they may learn that there are several reasons why sauces separate as well as a method of repairing separated sauces. Thus, when knowledgeable readers piece together familiar concepts (e.g., key ingredients of sauces, hot water) in novel ways (e.g., gradually add the separated sauce to a few tablespoons of hot water), they may take longer to integrate and organize these newly formed concepts than would readers without background knowledge.
The effects of knowledge on resource allocation do not differ for young and older adults in the present study. Knowledgeable older adults increase resource allocation to domain-related texts to the same extent as do their younger counterparts. These data suggest that older adults are capable of increasing effort during knowledge application.
In contrast to the findings reported here, some past research has shown that knowledge reduces encoding time (Miller & Stine-Morrow, 1998; Sharkey & Sharkey, 1987). This apparent contradiction could be reconciled by taking into account the match between the reader's knowledge and the text. When the match between the knowledge base and the text is high, as was the case when schemas were used in Sharkey and Sharkey (1987) and Miller and Stine-Morrow (1998), presumably there is a large overlap between concepts (and relations among concepts) presented in the text and the reader's knowledge base, which may enable knowledge to be applied rapidly and without effort. Conversely, when the match between the text and the reader's knowledge base is lower (e.g., when texts do not depict commonly held schema, when the text is low in coherence or contains information that is novel, ambiguous, contradictory) as may be the case in the present study, retrieval structures provided by knowledge appear to require greater effort to construct. Without ready-made connections, more time may be needed to organize and integrate information within the text as well as to organize and integrate the text with prior knowledge.
Knowledge and Memory
The effects of knowledge on recall of domain-related passages are consistent with a good deal of research showing that knowledge (e.g., Spilich et al., 1979; Voss et al., 1980) and expertise (e.g., Arbuckle et al., 1990; Morrow et al., 1992) enhance memory for text. The magnitude of this enhancement is modest in the present study (HK readers recall 18% more of domain-related texts than do LK readers) compared with studies in which aviation experts (23% more; Morrow et al., 1992) or music experts (28% more; Arbuckle et al., 1990) were tested or studies in which extreme groups of baseball knowledge were compared and individuals with moderate levels of knowledge were omitted (35% and 50% more for Spilich et al., 1979, and Voss et al., 1980, respectively). Thus, given that participants in the present study were not selected based on their experience in or knowledge of cooking, the data from the present study are encouraging because they indicate that even at relatively low levels on the expertise continuum, knowledge enhances memory for text.
The beneficial effects of knowledge are comparable for both age groups. The finding of age constancy in the effects of knowledge on memory is consistent with some past research (e.g., Arbuckle et al., 1990; Morrow et al., 1992) but inconsistent with the notion that older adults use knowledge to offset age-related declines in memory for text (e.g., Dixon & Bäckman, 1993). Although it is unclear why knowledge failed to reduce age differences in memory performance, it could be that the criterion task is not close enough to the domain in question. Morrow, Leirer, Fitzsimmons, and Altieri (1994) have argued that the extent of mitigation increases as the criterion task becomes more salient to the domain of expertise. Thus, having cooks read and recall recipes rather than scientific cooking articles may produce stronger knowledge effects on memory performance of older readers than those reported here.
Because HK readers in the present study also allocate more time to conceptual processing during encoding, these data suggest that the benefits of knowledge on performance may be due to increased effort. Past research is consistent with this notion. For example, McNamara, Kintsch, Songer, and Kintsch (1996) showed that HK readers gained more from their knowledge when texts were minimally coherent relative to when texts were fully coherent. Minimally coherent texts presumably required readers to allocate additional time to fill in the gaps that were not explicitly mentioned in the text. Although online measures were not included in that study, the data suggest that the advantages of domain knowledge on comprehension and memory may be attributable to increased effort during encoding.
In general, the data from the present study show that both younger and older knowledgeable readers increase effort when reading domain-related texts. This time probably paid off in terms of memory performance in that knowledge also increases memory for the domain-related texts. These data suggest that although the use of knowledge during reading can be effortful, older adults apply similar amounts of effort as do younger adults and have comparable gains in memory performance.
|
Appendix |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
|
Acknowledgments |
|---|
Portions of this research were presented at the Annual Convention of the American Psychological Association, Washington, DC, August 2000.
I am deeply indebted to the many culinary experts who participated in this research and in particular to those at the American Institute of Wine and Food, the Cambridge School of Culinary Arts, and the Culinary Guild of New England. Many thanks are also due to Rebecca Yun for her assistance in data collection and to Brian Miller for his unending support. I also wish to thank Liz Stine-Morrow, Dan Morrow, and Chris Herztog, who provided helpful comments on an earlier draft of the article.
|
Footnotes |
|---|
Received for publication August 6, 2001. Accepted for publication March 1, 2003.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. D. Castel Aging and Memory for Numerical Information: The Role of Specificity and Expertise in Associative Memory J. Gerontol. B. Psychol. Sci. Soc. Sci., May 1, 2007; 62(3): P194 - P196. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
