Adaptation of Clinical Evaluation of Language Functions – 4 th Edition to Brazilian Portuguese

Purposes: To translate and adapt the Clinical Evaluation of Language Functions – 4 Edition (CELF-4) to Brazilian Portuguese. Method: One hundred and sixty normal language development school children between the ages of seven and ten, half from public schools and the other half from private schools, both located on the east side of São Paulo. Results: CELF-4’s translation and adjustment to Brazilian Portuguese language showed equivalence between the original and translated versions, which demonstrates that there were no significant changes in the test’s form and content. Cronbach’s α test was used in order to verify CELF-4’s subtests internal consistency, in other words, if every subtest measures consistently the evaluated constructors. In this analysis, we observed that by excluding right or wrong items, and problematic items from the pool (those different from the rest of the group), all analyzed subtest presented satisfactory internal consistency, except for the Word Association Task for eight years old. Conclusion: Most subtests, as well as the Pragmatic Profile and the Observational Evaluation Scale, were simply translated, dismissing significant adaptations. The alterations performed were due to morphosyntactic and phonological differences between both languages. CELF-4’s translated and adapted version to Brazilian Portuguese was able to characterize the language performance in the studied population.


INtRODUCtION
The purpose of language assessment processes is to collect reliable information that is to be integrated and interpreted to make a judgment or a decision.This information contains the clinical history of the individual assessed, results of formal tests, and information collected from observations of natural situations (1)(2)(3)(4) .
For the assessment process to be effective, it is necessary to use adequate instruments and procedures to verify the language patterns of the child in question.These instruments must meet the conditions of the age range and language possibilities, as well as enable the best possible access to the child's linguistic potential (16) .
In Brazil, only two tests that assess children's language are available on the market: the ABFW Child Language Test (17) and the Protocol of Behavioral Observation (18) .As pointed out by many authors, it is possible to observe the scarcity of formal tests to assess children's language in Brazil (16,17,(19)(20)(21) .Some authors suggest that the translation and adaptation of instruments available into other languages can assuage this paucity.Moreover, research studies with this purpose can propitiate transcultural studies that compare Brazilian to international findings.In this manner, study objects would be better characterized in the sense that several research centers would use the same instruments (22)(23)(24)(25) .
Recently, in Brazil, some studies have been carried out with the purpose of translating and adapting formal tests of children's language into Brazilian Portuguese (BP) (21,26) .Although these studies have been completed, difficulties are still found in the country to diagnose children with language alterations and also to conduct research in this area, due to the lack of reliable reference standards about the linguistic abilities of children undergoing normal language development, especially for age ranges that surpass 7 years.

PURPOsE
The objective of this study was to translate and adapt the Clinical Evaluation of Language Functions -4th Edition (CELF-4) (27) into BP.

Ethical aspects
This study was analyzed and approved by the ethics committee of the Clinics Hospital of Universidade de São Paulo's School of Medicine (approval report number 145/10).
It is important to highlight that, before elaborating the project, we contacted Pearson Books, which published the CELF-4 (27) , with the purpose of obtaining authorization to translate the instrument into BP.The publishing house and the authors of the test gave us the permission to translate it and apply it only for the purposes of this research study.
In addition, all the individuals recruited agreed to participate in the study by signing an informed consent form.

sample
The participants were 160 students undergoing normal language development, ranging from 7 to 10 years of age, recruited from public and private schools in the eastern region of the city of São Paulo.The individuals were divided in four age ranges, totaling 40 individuals per age range, 20 from public schools and 20 from private schools.
The inclusion criteria were absence of language-related complaints or previous speech therapy, and satisfactory school performance, as reported by their teachers.Furthermore, they were required to achieve an adequate performance on a phonology test (28) and on the School Performance Test (SPT) (29) , used to ensure that the individuals included in the study had no phonological alterations and achieved adequate performances on reading, writing, and arithmetic tasks.All these elements, along with good school performances, guaranteed that the children were undergoing typical language development.The choice for the aforementioned tests was made considering the age range of the group studied.The children had to present performances that met the expectations for their age range and schooling, based on the parameters of each evaluation and on the phonology test, and to fall within the reference standards for their schooling on the SPT.Considering that authors of previous studies (21,26) did not find differences between the sexes, this variable was not considered.

Description of the test
The CELF-4, elaborated by Elisabeth Wiig, Wayne Secord, and Eleanor Semel in 2003, is an instrument used to assess communication and language disorders among students ranging from 5 to 21 years of age.According to the instructions manual, it must be administered individually; as it occurred in our study.
The instrument, which is administered in approximately 1 hour and 30 minutes, has previous editions and is available for speakers of English and Spanish.
The purposes of the test are to identify the existence of language disorders, to describe the nature of the language disorder, to evaluate abilities that underlie language (work memory, speech automation, and phonological awareness), and to evaluate contextual language and communication.
The CELF-4 test is composed of 2 image albums and 2 answer sheets, one for ages 5-8, and another for ages 9-21.
The instrument has 16 subtests that evaluate language subsystems and pragmatic profiles with a scale for observational evaluation, as shown in Chart 1.
The subtests that compose the CELF-4 are grouped in four levels of the evaluation process.Levels 1, 2, and 3 form the constructs or different scores generated by the test, as shown in Chart 2.

Procedure
The process of translation and adaptation of the test was carried out as follows: • Literal translation of the CELF-4 test (American English to BP); • Back-translation; • Selection of the individuals and application of the CELF-4; • Analysis of theoretical, semantic, and cultural equivalencebefore the application of the test on the individuals selected for the study, the researcher, the person in charge of backtranslation (an experienced researcher in the area of children's language, with a PhD in Speech-Language Pathology and Audiology, and fluent in English), and the supervisor of the present study (an experienced researcher in the area of children's language, with a PhD in Speech-Language Pathology and Audiology) examined each of the items on the subtests that compose the test, as well as the material used to apply it (in this case, the albums), to verify potential sociocultural discrepancies both in the subtest items and in the graphic material.No discrepancies were found, mainly because the material in question is recommended for children and it is graphically simple.The necessary adjustments pertained to phonological and morphosyntactic aspects, as in any translation.These adjustments were made during the translation process, when the need for them became clear; this happened, for instance, when the words for the phonological awareness subtest were selected.It is important to clarify that the phonemes to be discriminated, as well as their position in a given word, were maintained.
The CELF-4 test, translated and adapted into BP, was administered, in a quiet room in their schools, to all the individuals who met the inclusion criteria.
The answers were corrected and interpreted in accordance with the instructions contained in the examiner's manual, translated into BP.
The results of the test were classified according to the criteria adopted in the original version.

REsULts
The CELF-4 (27) was translated without difficulties.The original format of the test was not altered, that is, all the items tested were kept along with all application instructions (start, scoring criteria, interpretation of results, and answer sheets).The majority of the subtests as well as the Pragmatic Profile and the Observational Evaluation Scale (OES) were only translated, as significant adaptations were not necessary.
Alterations were necessary in the case of the subtests of Word Structure, Semantic Relations, Sentence Assembly, Paragraph Comprehension, and Phonological Awareness.The changes made were mainly due to morphosyntactic and phonological differences between the English and the Portuguese languages.

Descriptive statistical analyses and measures of reliability
The data were initially analyzed by descriptive statistics and measures of reliability, with the purpose of verifying whether the items of each scale presented good indices of internal consistency.The indices are used to analyze whether a given scale measures the construct to be analyzed in a consistent manner, that is, whether the items of a scale are consistent and adequate for the purpose in question.Following part of the same procedures used to validate the original material in English, we adopted Cronbach's α (for each scale and for each age, separately) as the measure of internal consistency in this study.The option for this type of analysis is considered because this statistical treatment is capable of measuring the internal consistency of the test.There is an important difference between this analysis and the one used in the original version, namely that while we had the data of 40 individuals in this study, in the American material the minimum number is 200 children per age range, which makes the analysis more encompassing.This factor, combined to the fact that the majority of the answers analyzed are binary (correct answer = 1; error = 0), substantially reduces the values of Cronbach's α, which can slightly affect this analysis, albeit without invalidating.It is noteworthy that the purpose of this study was not to validate the CELF-4, but to translate it and adapt it to BP.Clearly, in a validation study the number of individuals must be at least four times larger.
The acceptable values of Cronbach's α for tests that measure abilities are generally equal to or higher than 0.70.Therefore, we conducted new analyses for the scales that initially presented a lower value than the one mentioned above and removed the problematic items, with the purpose of reaching the minimum cutoff point.This does not mean that these items must be necessarily removed from the test (because of the limitations previously mentioned about sample size and the binary nature of the variables).Instead, they may eventually aid in understanding what types of items can be problematic for the sample that spoke BP in this study, and also indicate which items require special attention at the time of validating the translation and adaption to BP presented here.
The interpretation of a "problematic item" does not pertain to the absolute difficulty of an item (whether it was easy or difficult for the children); instead, it has to do with whether it is at par with the level of difficulty of a subtest, that is, if it seems to be discrepant in relation to the others.
Owing to the fact that the majority of subtests used binary data, the items that were very easy or very difficult (with correct or wrong answers provided by all the children) did not present any variance and were therefore eliminated, even from the initial analyses.
Upon exclusion of the items with correct answers given by the entire sample, as well as the problematic items, the subtests analyzed presented satisfactory internal consistency, except for Word Association in the case of the 8-year age range (Table 1).
The other subtests that compose the CELF-4 test presented Cronbach's α equal to or higher than 0.70, that is, a satisfactory internal consistency.

DIsCUssION
In order not to annul their use and applicability, the process of translation and adaptation of instruments published in other language must be thorough.Moreover, transcultural research studies are possible only if an equivalence between the original and the translated versions of the instrument is ensured.
In this study, the translation and adaptation of the CELF-4 followed a method already endorsed in other studies (30)(31)(32)(33) and recommended by the World Health Organization (34) .Thus, the steps followed were: (1) literal translation; (2) back-translation; (3) analysis of theoretical, semantic, and cultural equivalence; (4) test application; and (5) final version of the translation.
In this manner, the procedures followed for the translation and adaptation of the CELF-4 were compatible with the guidelines described in the literature with respect to the care demanded in the process of translating and adapting a foreign instrument.
While analyzing the equivalence between the translated and the original versions of the CELF-4, we came across difficulties concerning the items that evaluate morphosyntactic aspects of oral language.In the process of translating the items of the TELD-3 -Test of Early Language Development (11) , difficulties were found pertaining to those that investigated syntactic and morphological aspects of oral language (26) , similarly to the ones we encountered while translating the CELF-4.This seems obvious, considering the structural differences among languages, especially between Germanic languages (as the original language of both tests) and Latin languages (which is the case of BP).
Although changes were necessary in the items that evaluated phonological and morphosyntactic aspects, we can affirm that these adaptations were not significant, as they did not interfere with the internal consistency of the items evaluated and did not modify what the items of the test had the purpose of measuring, as proved by our results.
Considering that the analysis of the effectiveness of an instrument is part of its translation process, the translated instrument has to be administered in its new context for its validity to be analyzed.In the same manner, to consider a test as valid, it needs to be clinically accepted and be able to measure what it purports to evaluate with exactitude.
In this study, we used Cronbach's α to verify the internal consistency of each subtest of the CELF-4, that is, whether each subtest consistently measured the construct that it purports to evaluate.In this analysis, we observed that upon excluding the items answered correctly or wrongly by the entire sample, as well as the items considered problematic, all the subtests analyzed presented satisfactory internal consistency, except for Word Association in the 8-year age range.This aspect can be understood as an isolated finding or be attributed to the fact that selecting the students by age range reduced the sample, and also to the fact that only a few items compose this subtest.Larger samples, along with thorough analyses of each item that was excluded in the analysis of reliability, will aid in taking the decision to remove any subtests or subtest items that composed the CELF-4 from the version in BP.Only after applying the test on a large scale will we be able to analyze whether it is necessary to reformulate the translation and adaptation of the subtest items for the purpose of making it more appropriate to the Brazilian context.
At this point, we are not yet able to analyze the sensitivity and specificity of the CELF-4 (translated and adapted to BP), which are fundamental elements for its validation.For this purpose, the test needs to be applied on a large scale, that is, with a much larger number of individuals in different age ranges and from different regions of the country.
It is also worth highlighting that to analyze these parameters, it is important that the version of the CELF-4 translated and adapted to BP must applied on children with language alterations so as to verify its capability to detect alterations in the aspects of language that it purports to evaluate.

CONCLUsION
Our purpose in this study was to translate and adapt the CELF-4 (27) to BP.At the end of the study, we observed that: 1.The translation of the CELF-4 was carried out without difficulties that hindered its viability, and the entire original format of the test was maintained (start, scoring criteria, interpretation of the results, and answer sheets); Initial Cronbach: the items answered correctly by all individuals (x) were removed Final Cronbach: the items answered correctly by all individuals (x) and the problematic items ( ) were removed *Owing to the fact that there are only 3 items in this scale, values between 0.6 and 0.7 can be considered satisfactory.However, in the case of the 8-year age range, the scale seems poor **Values considered as satisfactory (a≥0.70)Caption: WA = Word Association 2. The majority of the subtests, as well as the Pragmatic Profile and the OES, were only translated and did not need any significant adaptations; 3. The translated and adapted BP version of the CELF-4 test proved itself adequate to characterize the language performance of the population studied in the abilities evaluated by the test; 4. The majority of the subtests presented satisfactory internal consistency, which means that the subtests that compose the CELF-4 indeed measure what they purport to evaluate; 5.The translated and adapted BP version of the CELF-4 must be applied on a large scale for the purpose of validating the test in BP.Its sensitivity and specificity must also be analyzed.
*ACPBG was responsible for data collection, tabulation and analysis as well as manuscript elaboration; DMBL was responsible for the project, study outline, and overall supervision of the stages of manuscript elaboration and writing.

Table 1 .
Analysis of the items (sum of scores) and of the internal consistency of the Word Association subtest*