Title:

Influence of educational level in aphasia testing: experiences from standardizing the Portuguese version of the AAT (Aachener Aphasie Test)

Author:

Martin Lauterbach, Neurologist and Psychiatrist, LEL (Language Research Laboratory), UNIC (Neurological Clinical Investigation Unit), Faculty of Medicine of Lisbon

e-mail: mlauterbach@fm.ul.pt

Postal address: Laboratório de Estudos de Linguagem, Centro de Estudos Egas Moniz, Hospital Santa Maria, 1649-028 Lisboa

Introduction

The AAT was developed, validated and standardized in the early eighties in the Department of Neurology of the University of Aachen. Since the publication of the original version in 1983 adaptation of the AAT to English, Dutch, Italian and Thai have been published. Standardizations of the original version for the Austrian and Swiss population have been elaborated. One can say that the AAT is now a days the most widely used aphasia battery in Europe.

In order to create a unitary diagnostic instrument to assess aphasic language impairment in Europe the translation, validation and standardization of the AAT for Portuguese was undertaken. The availability of such a unitary instrument could serve for the organization of multicenter studies. In studies to assess rehabilitation effects the centers of one country often cannot collect a sufficient number of patients fulfilling demanding inclusion criteria. For the purpose of cross-language, bilingual studies and the comparison of linguistic aspects in the symptoms of aphasia a unitary assessment tool is essential.

In Portugal in the last 30 years an adaptation of the MAE (Multilingual Aphasia Examination) was in use (Lit.) Its classificatory power was demonstrated in various studies and was compared to other aphasia batteries (Lit.).

The aim of this study is to describe the construction of the Portuguese version of the AAT (PAAT), its validation and standardization. The yardstick for a succeeded Portuguese version will be, if the translation maintains the psychometric properties of the original version and the already translated versions.

Structure of the PAAT

The Portuguese version of the AAT has the same formal structure as its original version. The PAAT consists of a semi-structured interview, the Token Test and four subtests which vary linguistic modalities and linguistic units in between the single parts. The translation considers the regularities of the Portuguese language, in order to vary the difficulty of the chosen elements. Each subtest is divided into between three and five parts, giving sixteen parts all together. Each part is ten items long.

The test starts with a semi-structured interview that is approximately 10 minutes long, in order to collect a sample of spontaneous language production. The interview is recorded and later on evaluated on a six point scale in six different levels of linguistic description.

The Token Test, an adaptation of Orgass (1986) of De Renzi and Vignolo (1962), comprises 5 parts of ten items each. The parts are ordered following increasing difficulty of the items.

The repetition subtest has five parts. Each part involves the repetition of a different unit of language. The units are (1) isolated sounds, (2) one to three syllabic, monomorphemic words with consonant clusters (e.g. mestre = master), (3) loanwords with untypical phonotactic for Portuguese speakers (e.g. cocktail), (4) compound words with increasing number of bound and alternately free morphemes (e.g. irresponsabilidade = irresponsibility for bound morphemes; marinheiro de água doce for free morphemes) and (5) sentences with increasing number of constituents.

The subtest written language is made up of three parts. The first part investigates reading aloud of words and phrases, the second part consists in composing words and phrases using letter and word tiles and in the third part the subject has to write to dictation. The items of the three parts are paralleled as close as possible in syllabic and phrase structure and also in frequency, in order to enable a comparison across the parts.

The naming subtest has 4 parts. All tasks require naming on confrontation of line-drawings, except part 2 where different colors are shown. In the first part common simple nouns from different semantic categories have to be named (e.g. mesa = table or cachimbo = pipe). The second part concerns color naming, which can be disturbed isolated in the case of alexia without agraphia. The third part concerns naming of compound words (e.g. saca-rolhas = corkscrew). In the forth part line-drawings of situations are shown, which has to be described by a phrase.

The comprehension subtest has for parts. Stimuli are presented in spoken or in written form. The subject answers by pointing to the target from a selection of four line drawings. One other of these drawings is a close semantic, phonemic or syntactic distracter. The other two drawings are not related to the target.

Adaptation into Portuguese

It was not possible to translate the material of the AAT from German to Portuguese in a simple one-to-one manner and preserve at the same time the underlying rationale for the construction of the test. In the Token Test the designation for circles círculo had to be changed into ball bola for subjects with less than 5 years of education, as círculo turned out to be to abstract for them. Bola is not the correct designation for a two dimensional object, but it is frequently used by lower educated people to designate a circle. In order to maintain the internal structure of the AAT, primarily the words for the subtests repetition, naming, written language and comprehension had to be substituted. For the respective subtests the Portuguese words were chosen to comprehend increasing complexity, decreasing frequency and increasing articulatory and lexical difficulty for native speakers of Portuguese in between each part. Whereas words of increasing length with regular CVCV structure represent a particular difficulty for speakers of German (e.g. Hepatitis), this is the common case in Portuguese vocabulary, as in romance languages in general. On the other hand consonant clusters pose a specific articulatory challenge for Portuguese (e.g. Volkswagen) which are highly familiar words for Germans. The formation of the compound word dishwasher in German is done by agglutination (Geschirrspülmaschine), whereas in Portuguese this word is formed by insertion of a preposition and sequencing of single words (máquina de lavar loiça). The first five items of the subtest naming of compound words are made up following this principle. Some compound words in Portuguese are composed of a noun with an inflected verb form (e.g. saca-rolhas = corkscrew). For the Portuguese version of the AAT the items 6-10 in the subtest naming of compound words employ such verb-noun combinations. This kind of words poses a particular difficulty for Broca aphasics, which tend to omit the verbal part of the word.

While in both, written and spoken, sentence comprehension parts the drawings from the Italian version of the AAT could be widely used the greatest part of the drawings for word comprehension had to be changed, in order to achieve the intended target-distracter relationship. Some of the items of the Italian version of the AAT could be used for the Portuguese version, e.g. autoestrada – estrada (port.) / autostrada – strada (ital.), what means highway – road; or escada rolante – escada (port.) / scala mobile – scala (ital.), what means escalator – stairs. For item-distracter pairs constructed on a homonym relationship (e.g. vela can mean candle or sail) or on rimes (fonte – ponte in English fountain – bridge) the drawings had to be created new.

All new drawings were tested in a pilot study with 60 healthy participants, in order to investigate their unanimous recognisability. Items that not fulfilled this criterion were excluded. On the following step the final assembling of the test material was done and standardization procedure of the Portuguese version of the AAT started.

Participants

The control group was composed of 153 healthy participants, stratified for age and years of education. Details of demographic composition are shown in Table 1

Table 1: Crosstabulation gender-age- education

sexo			years of schooling				Total
			illiterate	1-4 years of schooling	5-9 years of schooling	> 9 years of schooling
male	age-group	<=35	0	0	8	14	22
		36 - 64 years	0	8	9	7	24
		>=65 years	3	4	3	2	12
	Total		3	12	20	23	58
female	age-group	<=35	0	0	12	23	35
		36 - 64 years	0	14	11	17	42
		>=65 years	7	5	1	5	18
	Total		7	19	24	45	95
male and female	Total		10	31	44	68	153

Results

The data were first analyzed with two separate one-way ANOVAs for each of the 2 demographic variables:

age-group (1-3): 1: < 36 years; 2: 36 – 64 years; 3: > 64 years and

schooling group (0-3): 0: illiterate; 1: < 5 years of schooling; 2: 5 – 9 years of schooling; 3: > 9 years of schooling.

Furthermore we tested for sex differences using an independent-samples t-test. The dependent variables were the subtests of the PAAT. The subtest written language and the two parts, reading comprehension of words and reading comprehension of phrases, of the subtest comprehension were analyzed without the 10 illiterate participants, resulting in a subset of 143 participants.

Influence of gender

All comparisons for item-groups did not reach significant differences between male and females subjects, except the fourth part of the subtest comprehension. Only the part reading comprehension of phrase showed a significant difference between men and women (p = 0.036; mean male: 27,31 points, mean female: 28,27 points; maximum score of 30 points).

Influence of education

Table 2 shows the means scores and standard deviation of each education-group for the single subtest. The performance in the subtests improves as the years of schooling increase.

Table 2: mean scores and SD for education-groups

Subtest		Token Test	Repetition	Naming	auditive Compreh.	written language	compreh.	reading compreh.	compreh. of words	compreh. of phrases
max. score		50	150	120	60	90	120	60	60	60
illiterate N = 10
	Mean	41,30	139,70	94,80	47,5
	Std. Deviation	2,41	6,36	5,41	5,32
1-4 years of schooling N = 31
	Mean	44,19	148,55	108,81	54,35	85,68	106,71	52,35	54,35	52,35
	Std. Deviation	3,64	2,86	6,87	4,90	4,04	9,03	5,28	4,45	5,33
5-9 years of schooling N = 44
	Mean	48,02	149,93	114,55	56,82	88,48	112,07	55,23	57,07	54,99
	Std. Deviation	2,26	0,25	5,33	2,88	1,94	6,34	4,30	3,27	3,90
> 9 years of schooling N = 68
	Mean	48,37	149,94	117,04	59,34	89,63	117,66	58,41	58,90	58,85
	Std. Deviation	2,50	0,34	2,40	1,34	0,77	2,61	1,92	1,49	2,0
Total	N	153	153	153	153	143	143	143	143	143

Schooling yielded highly significant differences (p < 0,001) for all subtests. Subsequent multiple group comparisons using the Tamhane post-hoc Test yielded the following results, as shown in table 3 (Groups are ordered from left to right according to increasing mean performance. Age underlined groups showed no significant pair-wise differences):

Table 3: multiple pair wise comparisons of education groups, Tamhane post-hoc Test alpha < 0,05%, corrected by Holm procedure

Subtest	Number of participants	education-groups (0 – 3);
Token Test	153	0 1 2 3
Repetition	153	0 1 2 3
Written language	143	1 2 3
Naming	153	0 1 2 3
Comprehension	143	1 2 3

Influence of age

Table 4 shows the means scores and standard deviation of each age-group for the single subtests. The performance in the subtests decreases as the participants age increase.

Table 4: mean scores and SD for age-groups

Subtest		Token Test	Repetition	written language	Naming	compreh.	auditive compreh.	reading compreh.	compreh. of words	compreh. of phrases
max. score		50	150	90	120	120	60	60	60	60
=< 35 years of age N = 57
	Mean	48,86	149,95	89,42	116,79	116,65	58,88	57,86	58,44	58,30
	Std. Deviation	1,66	0,29	0,98	3,0	3,50	1,61	2,42	1,83	2,42
36 – 64 years of age N = 66
	Mean	46,92	149,88	88,30	114,32	112,85	57,30	55,54	57,10	55,76
	Std. Deviation	2,66	0,87	2,62	4,91	6,31	3,0	4,08	3,27	3,82
=> 65 years of age N = 30/20
	Mean	43,43	145,20	85,95	103,93	107,15	51,90	53,05	55,10	52,05
	Std. Deviation	4,69	5,87	4,27	9,96	11,83	6,51	6,95	5,58	6,86
Total	N	153	153	143	153	143	153	143	143	143

The ANOVA for age yielded highly significant differences (p < 0,001) for all subtests. Subsequent multiple group comparisons using the Tamhane post-hoc Test yielded the following results, as shown in Table 5, (again, groups are ordered from left to right according to increasing mean performance. Schooling groups in parentheses showed no significant pair-wise differences):

Table 5: multiple pair wise comparisons of age groups, Tamhane post-hoc Test alpha < 0,05%, corrected by Holm procedure

Subtest	Number of participants	age-groups (1 – 3)
Token Test	153	3 2 1
Repetition	153	3 2 1
Written language	143	3 2 1
Naming	153	3 2 1
Comprehension	143	3 2 1

Discussion

Gender differences were not very prominent. Only in the item-group reading comprehension of sentences women showed better performance than men. One can suppose that in dealing with more pragmatic aspects of language, women have an advantage over men.

For the other two demographic variables age and schooling we found significant group differences on all levels of language processing. There is a clear tendency that older (age-group 3) and lower educated (schooling-group 1) subjects differ significantly from the other groups. But also the better-educated and younger subjects showed some group differences.

The subtests written language, naming and comprehension were was the most sensible subtests with respect to education-group differences. All groups differed significantly from each other. For the written language subtest this is in some way a circular finding, as school teaches reading and writing. Nevertheless it was an unexpected finding that even for the quite simple tasks, like the PAAT subtest written language, the basic instruction of four years of schooling does not provide a ceiling effect in respect to literate abilities. For the naming and comprehension subtest pragmatic aspects and perceptual abilities that are modulated by education seem to exert influence on the test performance. This point is being addressed in the paragraph about naming.

The subtests “naming” and “Token Test” were the most sensible subtests with respect to age-group differences. The Token Test is known to be age sensitive. Already in the original version of the AAT a correction for age has been introduced. Naming capacity is known to decrease only from the seventh decade of life on. The fact that also the groups of younger participants differ significantly in naming performance point to that non-linguistic factors may influence the test performance.

All together these findings corroborate the opinion of Hawkins KA and Bender S, 2002 that age and educational level had to be taken into account, when interpreting neuropsychological testing. This precaution does not only concern the “naming”-subtests of language batteries. We found significant group differences in all subtests of the PAAT-battery. Consequently the Portuguese version of the AAT has to introduce correcting factors for age and educational level.

Looking at Table 1, one can notice that the composition of our sample is not balanced. It was difficult to include young, lower-educated and old better-educated subjects. This fact reflects on the one hand a demographic reality, but on the other hand one must have in mind that the population tested frequently has different characteristics. Often older, lower-educated subjects are overrepresented in patient-groups.

Naming (supplementary study)

Introduction: As noted above the naming subtest showed the strongest influence of the two demographic factors education and age. In the normative study for the Portuguese version of the Aachen Aphasia Test (PAAT) 10% of healthy subjects made errors in the naming of line drawings. Elderly individuals and subjects with a low educational level had an even higher error rate. We conducted a supplementary study in order to characterize the type of naming errors produced by healthy subjects and the influence of education and gender on such errors.

Method: 72 healthy elderly, (66 to 91 years old), had to name 20 items chosen from the naming tasks of the PAAT. Each subject was presented both, the line drawing and the photograph version of the items. Subjects were grouped according to literacy (illiterates, 1-4, 5-9, >9 years of schooling), and were balanced for gender (37 female and 35 male), age and sequence of presentation. MMSE was administered to exclude dementia.

Error rate and error type were the dependent variables. The error type was classified by an expert group (10 speech therapists and 10 linguistic students) into 7 different types: perceptively related, part-of the-whole, semantically related, circumscription, perseveration, non-related answer, no answer, according to the classification used by Nicholas(3).

Results:

Error rate analyses – Sequence (t-Student: p=0.16) and gender

(t-Student: p=0,61) had no significant effect on the naming performance. Significantly more errors occurred in the line drawing version of the naming task (Student`s t-test, p= 0.001). In both versions, photographs (p<0,001) and drawings (p<0,001), the ANOVA showed a main effect for the independent variable education. There was a significant correlation between error rate and education (correlation of Pearson: p<0.001).

Error type analyses: There were significantly more no answers for the drawings (χ2: p=0,01). 39,6% of the errors were classified as perceptual errors, 17,4% as semantically related, 15,9% as not related at all, 11,8% as circumscriptions, 11,5% as no-answers, 2,42% as perseverations and 1,3% as part-of the-whole.

Concerning the comparison of error type women made significantly more circumscriptions (Mann Whitney U: p<0,016) and semantically related errors (Mann Whitney U: p<0,007).

Discussion: There is no ceiling-effect in confrontation naming for healthy elderly subjects. This means that especially for these subjects education must be taken into account when scoring naming performance as well for healthy as for aphasic subjects. The overall error rate was 24,4%.

Only the illiterates and the group with 1-4 years of schooling showed an effect of stimuli material. For the two lowest educated groups the error rate was significantly higher for the drawing stimuli.

Of all errors observed 39,6% were perceptive ones. In these cases the origin for misnaming is not located in the linguistic system itself. 78,8% of all perceptive errors were made by the illiterates and the 1-4 years schooling group. In concordance with Nunes et al, we also found that formal instruction helps to decode two-dimensional representations and hence turns the access to the stored structural description more precise. In order to avoid perceptive errors, the stimuli material for confrontation naming had to be chosen very carefully. This all together corroborate the opinion that for the lower educated subjects it is better to use photographs than line-drawings in confrontation naming tasks. Reis et al. even defend the opinion to use three-dimensional objects.

References:

· Hawkins KA., Bender S.. Norms and the relationships of Boston Naming Test performance to vocabulary and education: a review. Aphasiology, 2002;16(12):1143-1153

· Huber, W., Poeck, K., Willmes, K.. The Aachen Aphasie Test. Advances in Neurology 1984; Vol. 42: Progress in Aphasiology, edit by F. C. Rose. Raven Press, New York.

· Kermin, H., Deloche, G., Metz-Lutz, M. N., Hannequin, D., Dordain, M., Perrier, D., Cardebat, D., Ferrand, I., Larroque, C., Naud, E., Pichard, B., Bunel, G. (1991). Short report: The effects of age, educational backgrround and sex on confrontation naming in normals; principles for testing naming hability. Aphasiology, (5), 579-582.

· Lauterbach, M., I.P. Martins,I.P. & Wilmes, K.. The influence of educational level on the performance in language testing. JINS, 2003, Vol. 9, No.4: 514-515

· Nicholas, M., Obler, L., Albert, M., Goodglass, H. (1985). Lexical retrieval in healthy aging. Cortex, (21), 595-606.

· Nunes, M., Guerreiro, M., Reis, A., Pavão-Martins, I. (2000). Naming and literacy: the influence of age. Journal of the International Neuropsychological Society, (6) 400.

· Reis, A., Guerreiro, M., Castro-Caldas, A. (1994). Influence of educational level of non-brain-damaged subjects on visual naming capacities. Journal of Clinical and Experimental Neuropsychology, (16), 939-942.