Age of Acquisition and naming performance in Frisian-Dutch bilingual speakers with Alzheimer’s dementia.
A
Frisian-Dutch picture naming study to investigate the retention of words and
their meanings in bilingual dementia
Wencke S. Veenstra
Department of Neurology, University Medical
Centre Groningen , The Netherlands
Nick
Miller
Speech and language sciences, University of Newcastle-upon-Tyne , United Kingdom
Mark Huisman
Department of Psychology, University
of Groningen , The Netherlands
In this study the effect of age of acquisition (AoA) on naming
performance was investigated in twenty-six Frisian-Dutch bilinguals suffering
from Alzheimer’s dementia (AD). An experimentally designed picture-naming task
and rated AoA measures were used to assess naming performance for both
languages. Consistent with findings from healthy speakers we expected AoA to
effect word retrieval in the present speakers. A significant effect of AoA on
naming performance was found in both languages, indicating that words acquired
early in life are better preserved and retrieved in mild to moderate AD. No
difference in languages was found, the bilingual Alzheimer patients showed
problems in lexical retrieval in both L1 and L2. However, our patients
demonstrated more inappropriate code switching within the Frisian setting (L1)
than within the Dutch setting (L2) but always chose the appropriate language in
conversation. A qualitative analysis of the error responses investigated the
nature and locus of naming deficits. No differences in overall error
distribution were found between languages for both early and late acquired
words. This naming study showed a significantly higher percentage of
semantically than visually related errors. These findings have important
implications for understanding problems in lexical retrieval in bilingual
Alzheimer patients and the development of more adequate language and
neuropsychological tests for bilingual individuals with neurological
impairment.
Key Words: Age of
acquisition, bilingualism, picture naming, dementia
1.
Introduction
Reviewing the literature, it becomes clear that language behaviour in bilingual individuals with Alzheimer’s dementia (AD) differs across their languages. Several studies have demonstrated differential impairment in the first language and/or the second language in individuals with AD (Obler & Albert, 1984; Hyltenstam & Stroud, 1989; De Santi et al., 1990). Increasing our knowledge of factors influencing the nature and pattern of language attrition in bilingual speakers with and without dementia raises possibilities and challenges from a range of angles. With the increasing longevity of large sections of the world’s population an increase in bilingual people with dementia is likely to represent a rising number of people presenting in clinics who require assessment and management of their language status. Population movements within several areas of the world also mean that the size of the global bilingual population is set to rise significantly. In parallel to this, the study of language acquisition and loss in bilingual speakers has over the past decades been shown to offer important insights into the development and processing of language in the human brain. It was with the dual aims of adding to both the clinical and theoretical fields that the present study of language loss in bilingual speakers with dementia was undertaken.
Only a small number of studies have focussed on the problems
occurring in lexical retrieval in bilingual AD patients. Given the fact that
the aging population is growing considerably, a rapid increase in the number of
patients affected by AD is predicted. Therefore it is expected that the
bilingual population with dementia will increase extremely, also taking into
account the increasing number of immigrations into
Within both the field of bilingualism and monolingual language processing an issue that has come to the fore concerns the influence of age of acquisition on a range of linguistic functions and processes including reaction times, retention, attrition. Examination of bilingual speakers who have acquired their languages at different ages enables a unique window of insight into the issue of effects and nature of age of acquisition.
The effects of certain word attributes on word naming and lexical decision have been widely investigated in word and picture naming tasks. Several theoretical accounts of word retrieval have considered that the frequency with which a word occurs in the language is an important determinant of the speed and accuracy with which that word can be recognized and responded to (Oldfield & Wingfield, 1965; Jescheniak & Levelt, 1994). It has been widely accepted that high-frequency words are recognized more rapidly and more accurately than low-frequency words. The classic picture naming study by Oldfield & Wingfield (1965) suggested a linear relationship between picture naming speed and frequency. Morrison, Ellis and Quinlan (1992) re-analysed the data from Oldfield and Wingfield and found that age-of-acquisition (AoA) is a significant independent determinant of object naming speed and that word frequency plays no independent role when its correlation with other variables, such as AoA and word length, is taken into account. This is in line with the previous findings by Caroll & White (1973b) and Gilhooly & Gilhooly (1979). AoA significantly correlates with both word frequency and word length. As stated by Morrison, Chappell & Ellis (1997), high-frequency words are learned earlier in life and tend to be short. Therefore, it is possible that previous reports of frequency effects in word naming and/or word length in word naming may in fact have been AoA effects. In a study of naming performance, Morrison et al. (1992) pointed out that naming speed was determined by rated AoA and word length. Additionally, Morrison & Ellis (1995) showed that there was a marked effect of rated AoA in naming matched on frequency but no frequency effect in naming matched on rated AoA, which means that AoA rather than frequency is the major determinant of naming speed. Studies of the performance on name retrieval among brain-injured aphasics have argued that word frequency is an important determinant of naming accuracy (Ellis, Miller & Sin, 1983; Kay & Ellis, 1987; and many others). Again, in these studies, investigators have failed to consider the possibility that AoA rather than word frequency may be the important predictor. However, there appears to be a substantial amount or evidence suggesting that, all other variables being equal, words learned early in life are retrieved faster than later acquired words in naming tasks (Carroll & White, 1973; Gilhooly & Gilhooly, 1979; Morrison, Ellis & Quinlan, 1992; Morrison & Ellis, 1995; Barry, Morrison & Ellis, 1997; Ellis & Morrison, 1998). And that early acquired words may be more resistant to the effects of some forms of brain injury than later acquired words (Hirsch & Ellis, 1994).
Carroll & White (1973a) were the first to obtain rated AoA measures. In a typical study, adults were asked to estimate the age at which they thought they had learned a specific word, either in spoken or in written form, on a 9- or 7-point rating scale which ranges from below 2 years of age to more than 13 years of age. Most picture naming studies and word recognition investigations involving AoA rely on the measures obtained by Gilhooly & Logie (1980). In the few studies in which new ratings are collected, including the one presented here, the new ratings are based on either the 7-point Gilhooly & Logie (1980) scale or the 9-point Carroll & White (1973b) revised scale, both of which are widely accepted as standard methods of measuring word learning age in the AoA literature (Morrison et al., 1997). Ideally, word learning age should be derived from a child language database, but due to the lack of such a database, researchers have relied on adult AoA ratings. However, there are good reasons to think that rated AoA measures are valid ratings of the age at which words are acquired. Age of acquisition ratings correlate highly with objective ratings of word learning age. Carroll & White (1973a) found a significant correlation between rated AoA and objective measures obtained by a normative study of when children are able to read words; Gilhooly & Gilhooly (1980) obtained a high correlation between rated AoA and the rank order of the words in the norms on the standardized Chrichton and the Mill Hill Vocabulary scales. More recently, Morrison et al. (1997) showed a correlation between children’s naming performance and adult AoA ratings for 297 pictorial nouns. Further evidence for the validity of rated AoA can be found in Gilhooly & Logie (1980), Walley & Metsala (1990, 1992) and Jorm (1991). Concluding, Morrison et al. (1997) have indicated that rated word learning age is both a reliable and a valid measure of age of acquisition. Hence, they suggest that researchers intending to examine AoA effects should use objective measure of AoA when available, but where not, adult ratings of word learning age provide a reliable and valid measure of real age of acquisition.
The
literature is equivocal on specifying the distinction between early and late
acquired words. For example, Ellis & Morrison (1998) studied AoA effects in
lexical retrieval using objective measures of word learning age. These obtained
normative data showed the age at which different object names enter the child’s
vocabulary (Morrison, Chappell & Ellis, 1997). To obtain those objective
measures a database of 280 children was split into a younger group, ranging in
age from 2.6 years to 7.11 years, and a older group ranging in age from 8 to
10.11 years. The objective AoA of a word name was rated as the age at which 75%
of children named the item correctly, with or without an initial sound cue. In
a previous study, Morrison & Ellis (1995) presented 4 experiments
investigating the contribution of word frequency and AoA effects on word naming
and lexical decision. In the first experiment they varied AoA while matching
the other variables. They proposed that high AoA items had values of 5.0 or
above on a scale of 1-7 ( 8-10 years or older), while low AoA items had a value
of 2.5 or below (2-4 years or below). Across the AoA literature there seems to
be a need of distinguishing early AoA from late AoA, but no clear-cut
definition has been used. Based on the existing literature on age of
acquisition and the distinctions made in the research as described above
(Morrison & Ellis, 1995; Morrison et
al. 1997), it appears to be that the average of 9 years (i.e. estimated AoA
of 8-10 years) is the cut-off age of distinguishing early from late acquired
words.
There seems to be no general consensus regarding early versus late word learning age. Morrison, et al. (1995, 1997) mentioned that a relatively late rated AOA was the one of 5 on a scale of 1-7 (estimated 8 years and older), whereas low AOA items had a value of 2.5 or below (i.e., estimated AoAs of 4 years or younger). Based on these findings and according to objective age of acquisition ratings taken from Morrison & Ellis (1998), in this study a cut-off point of 8.0 years was taken, thus late acquired words had an average value of 8.0 years or above, and early acquired words had an average value of below 8.0 years.
The aim of this study was, to investigate whether Age of Acquisition affects whether words (and their meanings) are retained in the bilingual individual with Alzheimer’s dementia. In this cross-sectional study, naming data of 26 early Frisian-Dutch bilingual Alzheimer patients were investigated. In particular, the effects of Age of Acquisition on the subjects’ naming abilities in both their first and their second language. Furthermore, different relative proficiency levels of object naming across the two languages were researched. Finally, there is a side issue according to the analysis of error responses in the naming tasks, concerning whether AD speakers produce more naming errors that are visually related to the target or predominantly linguistically related. The nature of the error responses across the demented subjects, for both Frisian and Dutch, will be taken into account. Object naming requires the successful operation of perceptual, semantic mand phonological processes (Ellis & Young, 1988), and a failure of any of those processes will result in problems with retrieval and production of the target name. As described earlier, the nature of the naming deficit of dementia and aging, whether perceptual or linguistic has been controversial. A qualitative analysis of the nature of error responses demonstrated by our bilingual AD patients, could help to draw a conclusion on the nature and locus of the naming deficits demonstrated in (bilingual) individuals with Alzheimer’s disease. However, it is previously described in the literature that problems with perceptual processes have been shown to influence naming errors, especially in the later stages of the disease (Huff et al. 1986; Hodges et al. 1991; Martin & Fedio, 1983). According to our subjects with dementia severity ranging from mild to moderate, we would expect to find more error responses, which are semantically related to the target word than visually related to the target word.
2. Methods
2.1. Subjects.
Twenty-six people
diagnosed as suffering from Alzheimer’s dementia participated in this study.
All of them had premorbidly been high proficiency bilinguals. They had acquired
Frisian as their native language, and learned their second language (i.e.
Dutch) from the age of six, in a school setting, an environment in which
Frisian and Dutch were used alternately. All patients were residents of
Nieuw-Toutenburg, a psycho-geriatric nursing home in Noord-Bergum,
The actual subject group consisted of five males and twenty-one females. All subjects were aged between 61 years and 96 years, with a mean age of 84.6 years. The subjects were not highly educated; most of them had only completed elementary school. Their mean years of formal education ranged from 3 to 13 years, with a mean of 6.8 years. They all used Dutch on a regular basis prior to the onset of dementia symptoms. However, in all cases, Frisian was the most dominant language throughout their lives. Five of the twenty-six knew languages other than Frisian and Dutch, but our testing was conducted only in these languages.
2.2. Task
The main data collection was through a picture-naming test. The experimental stimuli, 90 black-and-white drawings of objects, were taken from the European naming-test (EC-project, 1990-1992), which originally consisted of 391 items, all nouns. Two equivalent test batteries of 90 pictures were constructed, one for each language. These were further subdivided into three lists: A, B and C that were randomly ordered for each subject. Several measures were obtained for each picture and its name. Where ratings were available, those ratings were used. Others had to be obtained through pilot studies feeding into the main study.
Name agreement. Name agreement is a measure of how well subjects agree on the target name for a picture. Dutch measures of name agreement were available for all 391 pictured items. Those were obtained from two groups of healthy males and females, aged 60-75 with either less than 10 years, or more than 10 years formal education. Frisian measures of name agreement were obtained through a pilot study of healthy bilingual elderly subjects who were asked to name all the initial 391 pictures in both Frisian and Dutch. The final selection of pictures that were used in this study consisted of 90 stimuli with name agreement values ranging from 70% to 100% for both languages.
Word frequency. The CELEX database was consulted to obtain frequency values for Dutch words (Baayen, Piepenbrock & van Rijn, 1993). Frisian word frequency values were obtained from the Fryske Akademy, the main research institute for Frisian language and history. The three sets each had an equal estimated average of word frequency.
Word length. Word length was defined as the number of phonemes for both Frisian and Dutch names.
Age of acquisition. Rated age of acquisition norms were obtained using the 7-point rating scale by Gilhooly and Logie (1980). AoA ratings were collected from 17 healthy, high proficiency bilingual adults, males and females, aged 45-75 years (Veenstra 1999). All were native speakers of Frisian, and their second language was Dutch. These bilingual adults were presented with two lists comprising the 90 Frisian target words and their Dutch translation equivalents, and they were asked to rate when they believed they had learned a word, in either spoken or written form. The target words were split into two bands, early and late, according to the average ratings across subject groups. AoA ratings were averaged across the 17 subjects, and the mean score for the 90 experimental items was calculated for both languages.
2.3. Procedure.
The subjects were seen on two occasions: one testing session for Frisian (L1) and one for Dutch (L2). Testing took place in two separate sessions with an interval of three to seven days, to minimise learning/carry-over effects. Each patient was given 5 practice trials, using pictures that, like the pictures in the test proper, also showed above 70% name agreement. This was to familiarize the subject with the procedure before presentation of the 90 experimental items.
Subjects were asked to name the picture. If they failed to respond or indicated difficulty perceiving the meaning of the picture, a semantic cue was provided. If the subject was still unable to name the picture, a phonemic cue was given, which was the first sound of the target word. Using phonemic cues allowed us to determine whether the error was due to misidentification or to a lack of knowledge of the concept.
The patients were tested in Frisian and Dutch, in two separate sessions with an interval of three to seven days to avoid day-to-day fluctuation present in dementia. Testing was done by two different examiners, one monolingual Dutch speaker and one bilingual Frisian-Dutch speaker. The order of the two conditions was randomized across all subjects, this also to neutralize possible learning-effects. In both occasions, the interactions were entirely monolingual including greetings, conversation, instructions and the actual administration of the test. The monolingual examiner could easily say she did not understand what was said when she was spoken to in Frisian. In addition, she would ask for the Dutch translation equivalent of the Frisian word. However, the bilingual examiner could not possibly feign ignorance of Dutch, so she would repeat in Frisian what the patient said when the patient switched to Dutch.
The recorded testing sessions were transcribed in standard spelling of Frisian and Dutch, and scored according to their relation to the target word. A scoring system for the naming task was based on Cooper (1994). One additional category was added to cover the full range of responses made. All parts of responses were considered in arriving at scoring. Each set of responses was scored independently by two raters. Correct responses included the target word, synonyms as defined by the Frysk Wurdboek (Visser, 1992) and Van Dale, Woordenboek van het Hedendaags Nederlands (van Sterkenburg & Pijnenburg, 1984). In addition, self-corrected responses were classified as correct. Errors were classified as other language, cued or incorrect responses. Other language responses included the translation equivalents of the target word. Cued responses were those that were classified according to whether the target word was provided in the designated language after a semantic cue or a phonemic cue, or whether another language response was produced following a semantic cue or a phonemic cue. When a subject failed to produce the correct response after a cue had been provided, the response was coded as incorrect.
Incorrect responses were classified following a modified version of the categorization presented by Bayles & Tomoeda (1983), which was applied to bilingual AD patients by Cooper (1994). The codes for incorrect responses were either unrelated or related to the target word. The related responses were analyzed according to their visual similarity or linguistic similarity to the target item. Furthermore, the linguistically related responses were divided into phonemically related and semantically related, which was subdivided into seven categories. The unrelated responses were subdivided into unrelated errors, perseverations and no response.
3. Results
In order to shed light onto whether Age of Acquisition has an effect on word retrieval in bilingual Alzheimer patients, all responses produced in the Frisian and Dutch naming tests were assessed, following the procedure described earlier. Correct responses and self-corrected responses were calculated and are shown in Table 1.
Table 1: Mean number of corrected and self-corrected responses (with standard deviations and 95% confidence intervals) per target word in Frisian (L1) and Dutch (L2).
|
|
Correct responses |
|
Self-corrected responses |
||
|
|
Mean (SD) |
95% CI |
|
Mean |
95% CI |
|
Frisian (L1) Dutch (L2) |
11.61 (6.80) 11.68 (5.38) |
[10.19 , 13.04] [10.55 , 12.80] |
|
1.26 (1.40) 0.89 (1.02) |
[0.96 , 1.55] [0.67 , 1.10] |
Naming performance
A repeated-measures analysis of covariance was performed on the set of 90 pictured items to test the effect of language on correct and self corrected responses. The covariates included in the analyses are AoA, word frequency, and word length. The analysis showed a significant effect of language [F(1,86) = 10.76, p = 0.001] and significant effect of AoA [F(1,86) = 24.37, p < 0.001] on the number of correct responses. The mean numbers of correct responses, adjusted for the influence of the covariates are 10.76 [9.04 , 12.48] for L1 and 14.68 [12.96 , 16.36] for L2 (99% confidence intervals between parentheses). They are presented in Figure 1.
Figure 1: Mean number of items named correctly in L1 and L2.
Code switching
Also a repeated-measures analysis of covariance was performed to test the effect of language on code switching (using the number of other-language responses). The included covariates are AoA, word frequency, and word length. The analysis showed a significant effect of language [F(1,86) = 13.42, p < 0.001] and a significant effect of AoA [F(1,86) = 17.73, p < 0.001] on the number of other-language responses. More errors are made within the Frisian setting (mean 6.15 [4.49 , 7.81]) than within the Dutch setting (mean 1.94 [0.28 , 3.59]). The corrected means are presented in Figure 2.
Figure 2: Mean number of other-language responses for both L1 and L2.
Errors
A qualitative analysis of errors was undertaken. The data were split into different error categories mentioned above. The proportion of error types and the average number of errors per target words are outlined in Table 2. No difference in overall error distribution was found between languages for both early and late acquired words. However, for both language conditions the proportion of error responses in late AoA was higher than in early AoA. Naming errors were explained either perceptually or linguistically. The results indicate a significantly higher percentage of semantically than visually related errors in both language conditions.
Table 2: Classification and distribution of errors across Frisian (L1) and Dutch (L2) for early and late AoA.
|
|
Frisian (L1) |
|
Dutch (L2) |
||||||
|
Error type |
Example |
n |
Mean |
% |
|
Example |
n |
Mean |
% |
|
Visual Semantic Phonological Unrelated Perseveration Nonresponse |
Wetterfal ® ‘Sjaal’ Oaljefant ® ‘Ko’ Kroan ® ‘Troan’ Foarholle ® ‘Kleed’ |
90 90 90 90 90 90 |
1.18 4.52 0.16 0.87 0.07 0.63 |
4.5 17.4 0.6 3.3 0.3 2.4 |
|
Oog ® ‘Voetbal’ Peer ® ‘Appel’ Bijl ® ‘Vijl’ Stropdas ® ‘Stoel’ |
90 90 90 90 90 90 |
1.09 5.06 0.12 0.91 0.04 0.84 |
4.2 19.4 0.5 3.5 0.2 3.2 |
|
Early Late |
|
72 18 |
5.75 14.11 |
22.1 54.3 |
|
|
38 52 |
4.63 10.58 |
17.8 40.7 |
4. Discussion
The results of this picture naming study confirm previous reports of deficits in word finding in AD. Consistent with other studies on naming abilities in bilingual speakers with Alzheimer’s dementia (Cooper, 1994), we found that bilingual speakers with AD showed problems in lexical retrieval in both their first language (Frisian) and their second language (Dutch). Based on the literature it was assumed that AoA is the best predictor of object naming speed and naming accuracy in healthy young adults (Morrison et al., 1992). In addition, Hodgson & Ellis (1998) reported that age of acquisition and name agreement were independent predictors of naming accuracy in healthy elderly individuals. It was observed that older people perform better in retrieving early acquired words than late acquired words. Also, older subjects retrieved object names with a high name agreement faster and more accurately than object names with a low name agreement. However, in the present study name agreement was controlled for and AoA was the most important variable. We found a significant difference in naming performance of bilingual speakers with AD, between early and late acquired words in both language conditions.
Consistent with findings from healthy speakers we expected AoA to effect word retrieval in the present speakers. Rated AoA significantly effects naming performance in Frisian-Dutch bilingual speakers with Alzheimer’s dementia. found in both languages, indicating that words acquired early in life are better preserved and retrieved in mild to moderate AD.
No difference in languages was found, the bilingual Alzheimer patients showed problems in lexical retrieval in both L1 and L2. However, our patients demonstrated more inappropriate code switching within the Frisian setting (L1) than within the Dutch setting (L2) but always chose the appropriate language in conversation. These results are in line with the evidence presented by De Santi et al. (1990), who reported that their bilingual demented patient always chose the appropriate language in conversation but demonstrated code-switching problems in the first language. This pattern was explained by the fact that the patient had learned both languages almost at the same time, being an early bilingual. Also the subjects who participated in the present study were early bilinguals, as they started to learn their second language approximately at the age of six.
Hyltenstam & Stroud (1989) investigated two bilingual speakers with AD who had premorbidly been highly fluent bilinguals. Their results indicated that the late bilingual showed preference for speaking his first language, whereas the early bilingual did not show such preference for any of the two languages but did not always choose the appropriate language. According to Hyltenstam & Stroud the early bilingual suffers forma a language choice problem and the late bilingual suffers from a language separation problem of a disability to inhibit the native language while producing the second language.