Age-Related Differences in the Detection of Homophone Substitution Errors

Preliminary research suggests that older adults perform as well, if not better, than young adults on spelling detection tasks. This experiment investigated young and older adults’ detection accuracy of target words embedded in sentence contexts, contrary to previous studies that examined spelling detection of isolated words. Specifically, detection of homophone substitutions was examined by replacing a contextually appropriate word (e.g., beech) with its homophone foil (e.g., beach) or a pseudo-homophone (e.g., beetch). Age and spelling ability were investigated between participants; homophone dominance and the presence or absence of an orthographic prime were manipulated within participants. Spelling ability did not affect older adults’ detection of homophone substitutions, with both good and poor older spellers performing better than young adults. However, young good spellers were more accurate detectors than young poor spellers. Both age groups were less likely to detect errors when the dominant homophone replaced the subordinate homophone than when the subordinate replaced the dominant. Orthographic priming did not appear to influence detection of homophone errors for either age group. Results are explained using dual route models. Age-related changes in linguistic performance have been observed in both healthy older adults and those with age–related illnesses such Zoller: Age-Related Differences in Spelling Ability 229 as dementia (e.g., Abrams and Stanley 2004; Burke and Shafto 2004; MacKay, Abrams, and Pedroza 1999; Neils, Roeltegen, and Greer 1995). Longitudinal studies reveal age-related declines in the use of vocabulary and complex grammatical structures (Kemper, Thompson, and Marquis 2001). The language production literature shows older adults as having more difficulty than young adults in producing the proper spellings of words (MacKay and Abrams 1998) and in retrieving familiar words (Burke, MacKay, Worthey, and Wade 1991). Encouraging, however, is that older adults do appear to excel in some areas of language in comparison to their younger counterparts. Specifically, preliminary research suggests that older adults perform just as well, if not better, than young adults on spelling detection tasks (i.e., accurately identifying words as being correctly or incorrectly spelled) (MacKay, Abrams, and Pedroza 1999; Margolin and Abrams, in press). This research, however, has been limited to the detection of misspelled and correctly spelled words presented in isolation. The present study expands the research on spelling detection in older adults by investigating whether young and older adults differ in their ability to detect inappropriate spellings (i.e., homophone substitution errors) that are embedded in sentence contexts. Because people are rarely asked to detect a misspelling in isolation, spelling detection in sentence context allows for a more naturalistic observation of everyday errors made when proofreading written documents, letters, e-mails, etc. Research on age and spelling ability often compares young and older adults’ spelling detection with their spelling production. Indeed, older adults have been reported to have deficits in a variety of language production tasks, relative to young adults and relative to language comprehension or detection tasks. Typical language production tasks include word retrieval, picture naming, and spelling production tests. A common complaint among older people is the inability to produce familiar words. This “tip-of-the-tongue” phenomenon also occurs in young adults, but is much more prevalent among older adults (e.g., Burke et al., 1991). In particular, older adults report difficulties in retrieving proper nouns (e.g., names), especially those that have not been used recently or are not used frequently (Burke and Shafto 2004; Burke, Locantore, Austin, and Chae 2004). Age-related language production declines are also seen in picture naming tasks, where older 230 Chrestomathy: Volume 6, 2007 adults are asked to give the name of the object or person portrayed in a picture (Feyereisen 1997; Burke et al. 2004). Furthermore, there is evidence for age-related declines in spelling production (MacKay and Abrams 1998; Stuart-Hamilton and Rabbitt 1997). Spelling production in older adults is typically tested by asking participants to write down previously recorded words that are individually presented over computer speakers. Age-related declines seem to interact with word frequency, such that older adults are more likely to misspell common (high frequency) words than younger adults, but outperform young adults in the spelling of low frequency words (MacKay and Abrams 1998). The latter finding contradicts to the general consensus that older adults perform more poorly than young adults on production tasks. The enhanced performance by older adults of low-frequency production tasks may be explained by their heightened familiarity with low-frequency words in comparison to young adults. White, Zoller, and Abrams (2006) examined spelling production errors in a more naturalistic setting by asking young and older adults to transcribe entire sentences presented auditorily via loud speakers. We examined frequency and priming effects on spelling production of a homophone target word embedded in the sentence. The frequency effects observed were identical to those from MacKay and Abrams (1998): in relation to young adults, older adults were more likely to make homophone substitution errors with high frequency words, but less likely to make homophone substitution errors with low frequency words. Margolin and Abrams (in press) looked more specifically at agerelated declines in spelling ability as related to type of speller (i.e., “good” or “poor” speller based on percentage of correctly spelled words on a spelling test). For poor spellers, young adults performed better than older adults; however, among good spellers, young and older adults performed similarly. Categorization by speller-type seems to provide new insight into age-related deficits in spelling production. For older adults who are good spellers, performance on spelling tasks is very good and does not seem to be affected by increasing age. Among older adults who are poor spellers, however, poor performance on spelling tasks may indicate more potent age-related cognitive deficits for those who already perform poorly on cognitive tests. The age-related deficits found in much of the spelling production Zoller: Age-Related Differences in Spelling Ability 231 literature do not appear to generalize to tasks that involve spelling detection. Spelling detection involves the identification of correctlyspelled and misspelled written words. MacKay, Abrams, and Pedroza (1999) investigated spelling detection in young and older adults by briefly presenting correctly-spelled (e.g., endeavor) or incorrectly-spelled (e.g., endeavuor) words. Participants were told to judge if the spelling of the word was right or wrong. Older adults performed just as well as young adults on this task, and no significant age-frequency interaction was observed (see also MacKay and Abrams 1998). Abrams, White, McDermott, and Wolf (2000) tested spelling detection by giving young and older adults a multiple-choice recognition test in which they were asked to identify the correctlyspelled word among other possible spellings. Older adults identified spellings as being correct/incorrect more accurately than young adults. The Margolin and Abrams (in press) study discussed previously also investigated detection errors. Similar results were found for their detection and production tasks: increased age was detrimental only for poor spellers, and no age differences were found between young and older good spellers. Independent of spelling task, older adults were better than young adults at detecting pseudo-words (“fake” words imitating the phonology and orthography of real words, but having no meaning, e.g., beese, nooth), but performed similarly when detecting correctly-spelled and misspelled words. Thus, individual spelling ability was more predictive of age-related spelling differences than was type of spelling task. The spelling detection studies described above all presented participants with isolated words that were either correctly or incorrectly spelled. No published research has investigated older adults’ ability to detect misspellings in context, within sentences. Studying spelling errors in context provides a more naturalistic setting for evaluating the kinds of detection errors made in everyday life. Proofreading experiments employ such a naturalistic setting. Given a story passage, participants are instructed to read for meaning and cross out words that do not make sense in context. Each passage contains a homophone (e.g., beach), or a word that shares its sound (phonology) with another word (beech) that has a different spelling and meaning. Within a passage, correct (target) words are replaced with a spelling 232 Chrestomathy: Volume 6, 2007 control (a word spelled similarly to the correct word, e.g., batch), a homophone foil (the contextually inappropriate homophone, beach, replaces the contextually appropriate homophone beech), or a pseudohomophone (beetch replaces beach) (Jared, Levy, and Rayner 1999). The use of homophone and spelling controls in proofreading experiments provides an opportunity to examine how phonology and orthography affect performance in spelling detection tasks, something that has not been investigated with older adults. An elevated incidence of detection errors (identifying a word as being contextually appropriate when it is actually inappropriate, or vice versa) for homophones in comparison to spelling controls would provide evidence that phonology is used to activate word meaning (Van Orden, Pennington, and Stone 1990). Understanding the role of phonology in activating word meanings gives insight into developing methods that will improve recall of words, particularly for older adults. Much of the evidence from proofreading experiments with younger adults suggests that the phonology of a word is activated even before the meaning of the word is activated, especially for less skilled readers (Daneman and Stainton 1991; Treiman, Freyd, and Baron 1983; Van Orden 1987). Therefore, if a contextually inappropriate homophone (beach) stands in the place of the appropriate homophone (beech), a person is likely to leave the error undetected because the phonology of the word is activated and accepted (beach sounds like beech). These proofreading studies focus on young adults, and do not examine whether or not performance differs with age. Some evidence suggests that more skilled readers are less likely to activate phonology prior to activating a word’s meaning; rather, they are able to go directly from the printed word to its meaning (Jared and Seidenberg 1991). As one ages and gains more experience reading, one may be more likely to bypass a word’s phonology when activating its meaning. In such a case, older adults would be more likely to detect homophone substitution errors than young adults. This type of finding would contrast with language production studies (e.g., TOT) that generally show equivalent activation of phonology for young and older adults (James and Burke 2000; White and Abrams 2002). As mentioned above, a word’s meaning may be activated through a phonological route (e.g., Jared and Seidenberg 1991; Van Orden 1987; Zoller: Age-Related Differences in Spelling Ability 233 Van Orden, Pennington, and Stone 1990). In this case, a word’s phonology is first activated, and then semantic information is derived based on that phonology. For a homophone (beech), phonology is activated for both homophones (beech and beach). Thus, both word meanings are initially activated, and the appropriate meaning is then chosen based on sentence context. However, this phonological route is just one route to word meaning, as identified by dual route models (Van Orden 1987; Van Orden, Pennington, and Stone 1990). An alternative route to word meaning can also occur more directly, with word meanings activated directly from the printed word. The direct route is a more efficient method of meaning acquisition than the phonological route, and is believed to be the primary method by which good readers activate meaning from printed word (Jared and Seidenberg 1991). Because the phonological route uses a step-by-step process (spelling-sound-meaning) of activating word meaning, it is most useful for beginner and poor readers and when reading irregularly-spelled words (Jared et al. 1999). A correlation exists between reading and spelling ability, such that good readers tend to be good spellers and poor readers tend to be poor spellers (Kashner 1989). Good readers also have better vocabularies than poor readers (Kashner 1989). Therefore, it is logical that good spellers use the direct route as their primary means for activating words. Poor spellers would be more likely to use the phonological route. With presumably a 40-year advantage of reading experience over young adults, it seems logical that older adults would be more likely use the direct route for activating word meanings. Whether this experience advantage is enough to enable older poor spellers to also use a direct route remains to be seen. Word frequency is another variable that has often been manipulated in these proofreading studies. Jared et al.’s (1999) Experiment 2 asked young adults to read a passage and cross out items that did not make sense. They were also given comprehension questions to make sure they were reading for meaning. Participants were better at detecting errors when a low frequency word (beech) replaced a high frequency word (beach) than when a high frequency word (beach) replaced a low frequency word (beech). They were also more likely to detect errors when the target (beech) was replaced with a spelling control (bench) than when the target word was replaced with the homophone foil 234 Chrestomathy: Volume 6, 2007 (beach). The present experiment also investigated frequency in order to see if older adults have similar detection patterns as young adults for high and low frequency homophones. Note that in the present experiment, we use the terms dominant and subordinate to refer to our homophones. Dominance is often closely related to frequency (Starr and Fleming 2001; White and Abrams 2004a); however, it is possible for two homophones to have similar cumulative frequencies, but for one homophone to be dominant over another according to normative studies (White and Abrams 2004a). For example, bail and bale have similar cumulative frequencies (Zeno, Ivens, Millard, and Duvvuri 1995), respectively 9 and 5 words per million, but bail is considered more dominant based on White and Abrams’ normative studies. Additionally, by using homophone dominance, we were able to choose homophones that have been normed on both young and older adults, something that frequency norms do not control. Consistent with previous proofreading studies, we predicted that participants would be less likely to detect errors when the contextually inappropriate dominant homophone (beach) stands in place of the contextually appropriate subordinate homophone (beech) than when the contextually inappropriate subordinate homophone (beech) replaces the contextually appropriate dominant homophone (beach). Although proofreading studies have determined that homophone detection is influenced by word frequency, we were interested in additional variables that might influence detection. Specifically, we investigated whether the detection of substitution errors would be affected by prior exposure to words containing similar orthography to the target word. This orthographic priming has been shown to influence the production of homophone substitution errors (White et al. 2006). The current study employs a modified version of the priming paradigm introduced by White et al. (2006). In their production study, prior to spelling a target word (e.g., beech), participants wrote a prime word (e.g., teacher) that overlapped in orthography with the (contextually inappropriate) homophone (beach) of the target. Using this methodology, priming increased the likelihood that participants misspelled the target word in both young and older adults. This result suggests that spelling production errors (beach rather than the correct word beech) is influenced by the recent presentation of a similar Zoller: Age-Related Differences in Spelling Ability 235 orthography (teacher). Furthermore, young and older adults are equally affected by this exposure to orthography. This preliminary evidence suggests that orthographic priming is not reduced with increasing age, similar to phonological priming which has been shown to facilitate resolution of TOTs for both young and older adults (James and Burke 2000; White and Abrams 2002). In our investigation of priming effects on detection by older adults, participants were exposed to words (e.g., teacher) overlapping in orthography and phonology with a contextually inappropriate homophone (beach). Following this prime by four to eight words was a target word that was either the contextually appropriate homophone (beech), a contextually inappropriate homophone foil (beach), or a pseudo-homophone (beetch). Using this technique, we were able to establish whether recent activation of related orthography influences the detection of substitution errors, as it does the production of errors. We predicted an influence of orthographic overlap [EA], such that exposure to teacher and then beach (in a sentence where beech is contextually appropriate) would lead to an increased likelihood of detection errors (beach would be not identified as an error in the sentence) for both young and older adults. As for the relationship between spelling ability, age, and performance on our detection task, evidence from Margolin and Abrams (in press) leads us to believe that older adults who are poor spellers will perform worse on detection tasks than young adults who are poor spellers. For good spellers, both young and older, we anticipate similar performances on the detection task. The dual route theory should provide an accurate model for the means by which good spellers activate meaning efficiently through the direct route, and poor spellers activate meaning through the phonological route.