Carroll, J., & Sapon, S. (1959/2000). Modern language aptitude test (MLAT): Manual. San Antonio, TX: The Psychological Corporation. Republished by Second ...
Native Language Predictors of Foreign Language Proficiency and Foreign Language Aptitude Richard L. Sparks College of Mount St. Joseph Cincinnati, Ohio
Jon Patton, Leonore Ganschow, & Nancy Humbach Miami (Ohio) University Oxford, Ohio
James Javorsky Oakland University Rochester, Michigan
Fifty-four students were tested at specific time intervals over 10 years to determine best native language (NL) predictors of oral and written foreign language (FL) proficiency and FL aptitude. All participants completed two years of Spanish, French, or German. Each was administered measures of NL literacy, oral language, and cognitive ability in elementary school. A measure of FL aptitude was administered at the beginning of ninth grade and FL proficiency was evaluated at the end of the lOth grade. Among the variables, NL literacy measures were the best predictors of FL proficiency, and NL achievement and general (verbal) intelligence were strong predictors of FL aptitude. Results suggest that indices of NL literacy as early as first grade are related to FL proficiency and FL aptitude nine and 10 years later. Findings provide strong support for connections between L1 and L2 skills, and for speculation that "lower level" skills in phonological processing are important for written language development and oral proficiency in a FL. Annals of Dyslexia, Vol. 56, No. 1, 2006 Copyright ©2006 by The International Dyslexia Association® ISSN 0736-9387
129
130
SPARKSETAL.
Key Words: Foreign language aptitude, native language, prediction, proficiency Foreign language researchers have conducted studies on predictors of foreign language (FL) achievement for a number of years now, and prediction models have changed over time. (For reviews of these studies, see, e.g., Ehrman & Oxford, 1995; Lett & O'Meara, 1990; Onwuegbuzie, Bailey, & Daley, 2000; Skehan, 1989; Sparks, Javorksy, Patton, & Ganschow, 1998; Spolsky, 1995). These prediction models have incorporated general intelligence (IQ), language aptitude, affective variables (e.g., anxiety, motivation, attitudes, self-concept), personality (e.g., competitive versus cooperative, interaction style), demographic factors (e.g., gender, age), learning strategies (e.g., self-reward, memory devices), learning styles (e.g., study preferences, learning modalities), and other variables. Overall, FL aptitude and language-related variables have been found to be strong predictors of FL proficiency and achievement (Skehan, 2002). In recent years, the relationship between native language (NL) skills and FL learning has been studied extensively (e.g., see Ganschow & Sparks, 2001; Geva, 2000; Wade-Woolley, 1999). However, few long-term studies have been conducted that follow students from their early elementary school years when they were learning to read their NL to the time at which they began the study of a FL. In the United States, these studies are difficult to conduct because most students do not begin the study of a FL until the ninth grade. In addition, only one previous study has investigated the role of early indices of language skills in the development on FL aptitude, a variable that has been examined for over 40 years by FL researchers. In the present study, the authors examine the effects of NL skills including NL literacy (reading, spelling), NL oral language (receptive vocabulary, listening comprehension), and general intelligence (IQ) on oral and written FL proficiency and FL aptitude. They followed three cohorts of students from their entry into first grade through the 10th grade, at which time the students had completed two years of high school FL courses. The students' scores on measures of NL literacy, oral language, and general intelligence examined at specific time intervals were used as predictor variables to examine their impact on the development of language aptitude on a FL aptitude test and FL proficiency, which included measures of reading, writing, speaking, and listening to a FL.
NATWE LANGUAGEPREDICTORS
131
In the next section of the paper, the authors review studies that have examined the role of NL learning on FL achievement. Then, they explain the concept of FL aptitude, describe FL aptitude tests, and review studies that have examined the relationship between NL learning skills and subsequent FL aptitude. NL L E A R N I N G A N D FL A C H I E V E M E N T For several years, Sparks and Ganschow have speculated that there is a strong relationship between NL and FL learning. In their Linguistic C o d i n g Differences H y p o t h e s i s (LCDH) (Sparks, 1995; Sparks, & Ganschow, 1993a, 1995a; Sparks, Ganschow, & Pohlman, 1989), they hypothesized that FL learning is built on NL skills; that is, an individual's skill in the NL c o m p o n e n t s - - p h o n o l o g i c a l / o r t h o g r a p h i c , syntactic, and semantic--serves as the foundation for successful FL learning. Further, they posited that both native and FL learning depend on basic language learning mechanisms, and that problems with one language skill are likely to have a negative effect on both language systems. Other researchers have also suggested that there are connections between NL and FL learning. For example, Dufva and Voeten (1999) report that "well developed native language [NL] literacy skills" are "linguistic prerequisites of skilled FL learning" (p. 330). In reading, Geva (2000) hypothesizes that a student's ability to reflect on and conceptualize the sounds of his or her NL is related not only to the ability to read words but also to comprehend the foreign (L2) language. In his Linguistic Interdependence Hypothesis, Cummins (1984) defines language proficiency mostly in terms of linguistic skills and other factors related to language learning. He also hypothesizes that transfer of native (L1) language skills to the foreign (L2) language will occur only if a student has achieved a strong level of proficiency in his or her NL skills. Researchers have provided support for speculation that there is a connection between native and FL learning. For example, Humes-Bartlo (1989) reported that faster language learners in a population of fifth grade children attending bilingual classes had more highly developed NL skills than the slow language learners. Olshtain, Shohamy, Kemp, and Chatow (1990) found that NL proficiency played the most important role in FL learning among a group of 11- to 12-year-old Hebrew-speaking students studying English. Among Dutch students studying
132
SPARKS ET AL.
English, Hulstijn and Bossers (1992) reported that more individual differences in FL learning can be accounted for by individual differences in students' NL skills. In the past 15 years, Sparks, Ganschow, and their colleagues have conducted an extensive series of studies to test their hypothesis with good and poor FL learners at the secondary and p o s t s e c o n d a r y levels. (For reviews of this research, see Ganschow & Sparks, 2001; Ganschow, Sparks, & Javorsky, 1998; Sparks, 1995.) In their studies, they have found that FL learners who have significantly stronger NL skills achieve higher levels of oral and written proficiency and FL achievement (classroom grades) than FL learners with weaker NL skills. They have also f o u n d that p h o n o l o g i c a l / o r t h o g r a p h i c (i.e., s o u n d and s o u n d / s y m b o l learning) difficulties in the NL, in particular, play an important role in distinguishing good and poor language learners (Ganschow & Sparks, 1995, 1996; Ganschow et al., 1991; Sparks & Ganschow, 1993b, 1995b, 1996, 2001; Sparks, Ganschow et al., 1998; Sparks et al., 1992a, 1992b). In two prediction studies conducted with high school students, Sparks and his colleagues found that the best predictors of FL achievement were variables related to NL learning and language aptitude on a FL aptitude test (Sparks, Ganschow, & Patton, 1995; Sparks, Ganschow et al., 1997). Considerable evidence for the importance of the phonological/ orthographic component has also been found in recent crosslinguistic studies (see, e.g., Comeau, Cormier, Grandmaison, & Lacroix, 1999;. Durguno~lu, Nagy, & Hancin-Bhatt, 1993; Durguno~lu & Oney, 1999; Gottardo, Yan, Siegel, & WadeWoolley, 2001; Koda, 1992, 1998; Lindsey, Manis, & Bailey, 2003). These researchers have found relationships between phonological/ orthographic tasks in the NL and subsequent FL achievement. (For an extended discussion of this topic, see Koda, 1992, 1998). For example, a s t u d y by Muljani, Koda, and Moates (1998) showed that word recognition development in the NL facilitates word recognition in the second language. Meschyan & Hernandez (2002) reported that n o n w o r d decoding skill in English was predictive of second language decoding, vocabulary, grammar, and reading comprehension skills in college-age adults. Studies conducted with bilingual readers also have demonstrated that principles of learning to read can be transferred across languages, especially if the two languages are cognates of the same system (e.g., according to the alphabetic principle) (Bialystok, Luk, & Kwan, 2005).
NATIVE LANGUAGE PREDICTORS
133
Recent studies have also reported evidence of a relationship between phonological memory and FL learning. For example, Gathercole and Thorn (1998) found that poor phonological memory for the sounds of new words is likely to be considerably poorer for words in an unfamiliar language than in the NL. Service (1992) and Service and Kohonen (1995) found that pseudoword repetition measured prior to beginning the study of a FL predicted English learning. Phonological memory has also been found to be predictive of specific FL learning skills such as vocabulary acquisition and oral and written language comprehension (Cheung, 1996; Papagno, Valentine, & Baddeley, 1991). In a related study, Dufva and Voeten (1999) followed 160 Finnish children from the first through the third grade when they started to learn English. They found that both NL literacy and phonological memory have positive effects on FL learning. For example, the results showed that students' level of word recognition in Finnish was the strongest predictor of FL learning in third grade, indicating a strong relationship between students' NL skills, especially in the phonological/orthographic component of language, and subsequent achievement in a FL. These findings lend support to Sparks' and Ganschow's LCDH. NL L E A R N I N G A N D FL A P T I T U D E One of the first prediction models investigated in the FL research literature was the cognitive model, characterized primarily by a s s u m p t i o n s about a specific a p t i t u d e for l e a r n i n g languages (Spolsky, 1995). Proponents of this model assert that there is a talent for learning languages that is (a) independent of intelligence, (b) not the result of prior learning, (c) stable over time, and (d) different from individual to individual (Skehan, 1991). In early studies of FL learning, some researchers speculated that general intelligence was one type of aptitude important for FL l e a r n i n g (e.g., see, Briggs, 1922; Jordan, 1925; Symonds, 1930). In his model of FL learning, Pimsleur (1966a) included verbal intelligence as measured by a student's NL vocabulary as a component of FL learning. In the 1960s, studies showed that IQ as measured by standardized intelligence tests was not predictive of FL learning and achievement (Gardner & Lambert, 1972). In his studies on FL learning, Carroll (1962, 1993) found that FL aptitude is a specialized construct independent of intelligence. More recently, Sasaki (1996) reported that scores on FL aptitude tests (i.e., the Japanese version of the
134
SPARKSETAL.
Modern Language Aptitude Test subtests) and intelligence tests showed some separation between the two areas. Factor analysis of the results showed that some language analytic aspects of aptitude were closely related to intelligence, whereas other aspects of language ability (i.e., phonetic coding ability and memory) were less closely related to intelligence. Historically, the tests most commonly used to measure language aptitude in order to predict FL learning skill are Carroll & S a p o n ' s M o d e r n L a n g u a g e A p t i t u d e Test (MLAT) (1959/2000), Pimsleur's Language Aptitude Battery (LAB) (1966b), and the Defense Language Aptitude Battery (DLAB) (Petersen & A1-Haik, 1976). According to Skehan (1986a, 1989), depending on instructional conditions, these test batteries have correlations with foreign language achievement in the range of .40 to .70. The MLAT is still a popular aptitude measure in research studies despite its outdated norms. Its authors identified four factors that have been found through factor analytic studies to be predictive of FL achievement: (1) phonemic coding, or sound-symbol associations; (2) grammatical sensitivity, or the ability to recognize grammatical functions of words in sentences structures; (3) inductive language learning ability, or the ability to infer or deduce the rules governing a set of language materials; and (4) rote learning ability, or the ability to learn associations between sounds and meanings rapidly and efficiently (Carroll, 1962, 1981, 1990). (A description of the MLAT and its subtests is provided in the Appendix.) In recent years, Skehan (2002) has reported that the study of FL a p t i t u d e has been d e e m p h a s i z e d by FL researchers. Aptitude tests have been criticized for their failure to take into account different strategies learners might use to acquire a language and for their lack of relevance to the communication approaches used in most language classrooms today (Ehrman & Oxford, 1995). Furthermore, Krashen (1981) has claimed that aptitude tests are limited because they predict learning only in formal (i.e., classroom) situations. However, Skehan (1986a, 1986b, 1989, 2002) argues that aptitude is relevant to both explicit, conventional teaching contexts and implicit learning situations. His research has indicated that language aptitude tests are effective predictors of FL achievement because they measure relevant linguistic skills used in learning language in all contexts, and draw on skills in using decontextualized language (i.e., use of language skills in the classroom). To the authors' knowledge, there has been only one reported study that has examined connections between students'
NATIVE LANGUAGEPREDICTORS
135
skills in first (native) language and FL aptitude some years later. Skehan (1986a, 1986b, 1989, 2002) described a follow-up study of a subset of students from the Bristol Language Project, a longitudinal study of first language development of 125 children. In this project, Wells (1981a, 1981b, 1985) found that a great deal of variation exists in the speed at which children acquire their first language. His results showed that the children's "scores on measures of oral language at successive age-points yield progressively stronger predictions of later educational attainment" (Wells, 1981b, p. 181). Skehan extended these studies by investigating possible connections between the students" first language development and their skill in learning a FL several years later in order to "investigate the origin of [language] aptitude" (Skehan, 1986a, p. 197). He tetsed 23 students from Wells' studies who were now 13 to 16 years old and in secondary school to determine whether the data available from their first language development were related to their learning of a FL (French, German). His results showed strong correlations between early indices of first language development and FL achievement, and also between first language development and FL aptitude. The findings also showed that the FL aptitude measures accounted for most of the success in FL achievement. Skehan concluded that children who "developed more quickly in their first lang u a g e . . , are also those who tend to have higher foreign language aptitude" (p. 198). In his study, Skehan also found that there were strong relationships between family literacy background indices (e.g., education of parents, literacy orientation of the home, and students' level of FL aptitude several years later). More recently, NL researchers have conducted studies on the effects of literacy activities in the home (i.e., the Home Literacy Model) (Senechal, LeFevre, Thomas, & Daley, 1998). These researchers have made several proposals about the connections between literacy in the home and subsequent literacy development. For example, they speculate that parents differ in the types of literacy activities they include at home; storybook exposure and parent teaching about literacy are differentially related to their children's language, early literacy, and phoneme awareness skills; and formal and informal literacy activities practiced at home are both directly and indirectly related to later literacy development (Senechal, 2006). A number of studies have supported the premises of the Home Literacy Model, and have found that literacy activities in the home are related to both first and second language literacy development (e.g., see Bus, van Ijzendoorn, & Pelligrini, 1995; Dickinson, MaCabe,
136
SPARKS ET AL.
Clark-Chiarelli, & Wolf, 2004). The results of studies using the Home Literacy Model are supportive of Skehan's findings from the Bristol Language Project. P U R P O S E S OF S T U D Y The purposes of the present study were to determine (a) the best NL predictors of oral (listening, speaking) and written (reading comprehension, writing) FL proficiency; (b) the best NL predictors of FL aptitude (on the MLAT); and (c) whether NL predictors of FL proficiency and FL aptitude differ. METHOD PARTICIPANTS Fifty-four students attending a large, middle class, rural public school district in the midwestern United States who had completed two years of Spanish, French, or German courses in the ninth and 10th grades and were tested at specific time intervals over a 10-year period participated in the study. Thirty of the students had taken Spanish, 14 had taken French, and 10 had taken German. All of the participants had completed their second year of the FL by the end of the 10th grade. The study started with 156 students when they began the first grade in this school district. By the beginning of the ninth grade, 101 of the 156 students continued to attend school in the district. Seventy-seven of the 101 students began the study of a FL in the ninth grade. Fifty-four of the 77 students chose to continue their participation in the study when they began their first year in a FL course. Parental permission was obtained for each participant. There were 25 male participants and 29 female participants. The mean age of the participants at the beginning of the first grade was 6 years, 9 months (age range 6 years, 3 months to 7 years, 11 months); at the end of the study 10 years later, the mean age of the participants was 16 years, 4 months (age range 15 years, 9 months to 16 years, 11 months)3
1 The authors extendspecialthanks to Ms. JackieBlairfor her crucial role as facilitator and schoolliaisonin this study. The study could not have been completedwithout her valuableassistance.
NATIVE LANGUAGEPREDICTORS
137
The a u t h o r s d i d not have the resources to test all first grade students enrolled in the district's e l e m e n t a r y school in a given year. However, they w a n t e d to obtain a s a m p l e of s t u d e n t s w h o w o u l d be likely to take FL courses in h i g h school, a n d h a d a range of academic a c h i e v e m e n t skills (e.g., good, average, a n d below average readers). With these considerations in m i n d , the authors chose to use a cohort model. The 54 participants from three different cohorts (1992-93, 1993-94, 1994-95) were chosen in the following manner. Each year in kindergarten, the school district a d m i n i s t e r e d a r e a d i n g readiness measure, the Test of R e a d i n g Readiness: Level K (Scott, F o r e s m a n a n d C o m p a n y , 1987) to all of its k i n d e r g a r t e n students. Based on their score on this m e a s u r e (raw scores r a n g e d f r o m 0-34) a n d the k i n d e r garten t e a c h e r ' s r e c o m m e n d a t i o n , s t u d e n t s w e r e a s s i g n e d by the school to o n e of three t y p e s of first g r a d e classrooms: (1) above average readers, (2) at-risk readers, a n d (3) average readers. There w a s only one classroom of above average readers a n d o n e c l a s s r o o m of at-risk r e a d e r s e a c h year. 2 H o w e v e r , t h e r e were several classrooms consisting of average readers. For each of the three cohorts, the authors invited all of the s t u d e n t s in the above a v e r a g e classroom a n d the at-risk classroom to participate in the study. O n the Test of R e a d i n g Readiness, the above average s t u d e n t s h a d achieved r a w scores r a n g i n g from 33-34 a n d the at-risk s t u d e n t s h a d achieved r a w scores r a n g i n g fromJ 0-16. C o m p a r a b l e n u m b e r s of s t u d e n t s in the average reading c l a s s r o o m s w h o s e score o n t h e r e a d i n g r e a d i n e s s m e a s u r e r a n g e d from 17-33 were selected r a n d o m l y a n d invited to participate in the study. Of the 54 s t u d e n t s w h o participated over all 10 years of the study, 18 (33.3%) w e r e above average readers, 19 (35.2%) w e r e a v e r a g e readers, a n d 17 (31.5%) w e r e at-risk readers. There were similar n u m b e r s of s t u d e n t s in each of the three cohorts (1992-93 cohort = 17 s t u d e n t s , 1993-94 cohort = 20 students, 1994-95 cohort = 17 students). INSTRUMENTS
P r e d i c t o r Variables. There w e r e eight NL p r e d i c t o r variables in this study. Each of the variables is listed here followed by an acronym that is u s e d to represent the variable t h r o u g h o u t the r e m a i n d e r of t h i s p a p e r : (a) W o o d c o c k R e a d i n g M a s t e r y
2. In the third cohort, the school district did not separate the above average readers into a separate classroom. Some of the at-risk readers remained in the at-risk class through third grade.
138
SPARKS ET AL.
Test-Revised (WRMT-R) (Woodcock, 1987); (b) Test of Written Spelling-2 (TWS) (Larsen & Hammill, 1986); (c) Formal Reading I n v e n t o r y (FRI) (Wiederholt, 1986); (d) P e a b o d y Picture Vocabulary Test-Revised (PPVT-R) (Dunn & Dunn, 1981); (e) Lindamood Auditory Conceptualization Test (LAC) (Lindamood & Lindamood, 1979); (f) Test of Reading Readiness (READINESS) (Scott Foresman and Company, 1987); (g) Test of Cognitive Skills (TCS) (CTB/McGraw-Hill, 1983); and (h) listening comprehension assessed by the alternate form of the Woodcock Reading Mastery Test-Revised, Passage C o m p r e h e n s i o n Subtest (LCOMP). (See Aaron, 1989 for a complete description of the listening comprehension testing procedure). A description of each variable is provided in the Appendix. The aforementioned tests were chosen as predictor variables because each measured different components of language learning (i.e., phonological, phonological/orthographic, syntactic, semantic). Phonological skills, specifically phonemic awareness, were assessed by the LAC. Phonological/orthographic skills were assessed by the READINESS measure and the TWS. Phonological/orthographic, syntactic, and semantic skills were assessed by the WRMT-R and the FRI. Syntactic and semantic skills were assessed by LCOMP (i.e., WRMT-R Passage Comprehension subtest). Semantic skills were assessed by the PPVT-R. The TCS was used because its four subtests provided an overall measure of general intellectual ability (i.e., IQ). Foreign Language Proficiency Measures. The FL proficiency measures used in this study tested students' skills in the four areas identified by the American Council on the Teaching of Foreign Language (ACTFL) Proficiency Guidelines (1986, 1989) as essential for FL acquisition: reading, writing, listening, and speaking. Three FL educators, who were formally trained in ACTFL guidelines, developed the proficiency tests in their respective languages. The FL educators worked together to design parallel tests across the three languages (i.e., Spanish, French, German); that is, reading, writing, listening, and speaking in Spanish were tested by using the same items and prompts as those used for French and German. The only differences in the three reading comprehension tests were those specific to a particular FL. The test directions were the same for each of the three FLs. All of these proficiency measures had been used by the authors in their previous studies. (See Sparks et al., 1997; Sparks, Ganschow, Artzer et al., 1998 for a description of the FL proficiency measures.) A brief explanation of the measures is provided here.
NATIVE LANCUACEPREDICTORS
139
The FL reading c o m p r e h e n s i o n test was d e s i g n e d u s i n g criteria descriptive of the i n t e r m e d i a t e - h i g h level of t h e ACTFL guidelines. S t u d e n t s read a o n e - p a g e letter w r i t t e n in the FL and a n s w e r e d 10 multiple-choice questions in English a b o u t the letter.3 Then, the s t u d e n t s w e r e g i v e n a p a s s a g e to read f r o m Reader's Digest in either Spanish, French, or G e r m a n , a n d w e r e asked to a n s w e r 10 multiple-choice questions in English a b o u t the passage. The s t u d e n t s were given 15 m i n u t e s to read the letter a n d a n s w e r the questions, a n d 15 m i n u t e s to read the passage a n d a n s w e r the questions ( m a x i m u m score = 20). On the FL writing test, s t u d e n t s wrote a response in the FL to the letter that h a d been read for the r e a d i n g c o m p r e h e n s i o n task. The letter contained five questions to w h i c h the s t u d e n t s were asked to respond. The s t u d e n t s were given 15 m i n u t e s to write the letter. Each s t u d e n t ' s w r i t i n g s a m p l e w a s scored for five criteria: vocabulary, c u l t u r a l a p p r o p r i a t e n e s s , s t r u c t u r e , comprehensibility, a n d spelling. The ACTFL guidelines for det e r m i n i n g proficiency levels w e r e u s e d in assigning the scores (0-5) o n each of the a f o r e m e n t i o n e d criteria in the f o l l o w i n g manner. The s t u d e n t w a s evaluated in each area u s i n g a 5-point scale: 0 = no p r o d u c t i o n , 1 = novice-low, 2 = n o v i c e - m i d , 3 = novice-high, 4 = i n t e r m e d i a t e - l o w , 5 = i n t e r m e d i a t e - h i g h a n d above ( m a x i m u m score = 25). A score of zero w a s i n c l u d e d in the scoring because s t u d e n t s at this level of e d u c a t i o n m a y h a v e been unable to p r o d u c e a n y response in the target language. O n the FL listening a n d s p e a k i n g tests, students' proficiency w a s a s s e s s e d t h r o u g h a 10-15 m i n u t e i n d i v i d u a l oral proficiency interview. All of the interviewers w e r e at the S u p e r i o r level on the ACTFL OPI guidelines. The i n t e r v i e w e r s h a d no previous i n f o r m a t i o n a b o u t the participants, their NL or FL apt i t u d e test scores, or their p e r f o r m a n c e in the FL c l a s s r o o m . I n t e r v i e w e r s u s e d r a n d o m l y selected topics a b o u t w h i c h the s t u d e n t s conversed (e.g., family, food, school, friends, daily activities). Each s t u d e n t ' s interview w a s t a p e d so that it c o u l d be scored at a later date. 4 The oral i n t e r v i e w w a s scored for five criteria: p r o n u n c i a t i o n , vocabulary, g r a m m a r , c o m p r e h e n s i b i l ity, a n d listening c o m p r e h e n s i o n . The r e m a i n d e r of the scoring p r o c e d u r e a n d m a x i m u m score (25) w a s the s a m e as in the FL writing test. 3. The comprehension questions were written in English to conform to the ACTFL proficiency guidelines. 4. The authors thank Kelly Noe and Johannes Tokarski for administering the oral proficiency interviews to the second and third cohorts.
140
SPARKS ET AL.
The students' total proficiency score (FL TOTAL PROF) was the total n u m b e r of points on the FL reading c o m p r e h e n s i o n , w r i t i n g , a n d l i s t e n i n g / s p e a k i n g tests ( m a x i m u m score = 70). The justification for c o m b i n i n g the scores on the three tests is that the ACTFL guidelines define proficiency as reading, writing, speaking, a n d listening to a FL. The total s a m p l e achieved a m e a n total proficiency score of 25.2 (SD = 12.1) on the test. The students' scores r a n g e d from 5-58. The reliability of the FL TOTAL PROF was checked by calculating Cronbach's alpha. For the total test, the reliability coefficient w a s .86. Foreign Language Aptitude. T h e M o d e r n L a n g u a g e A p t i t u d e Test (MLAT) (Carroll & Sapon, 1959/2000) was u s e d to m e a s u r e students' FL aptitude. This test is d e s i g n e d to provide an indication of a s t u d e n t ' s probable degree of success in l e a r n i n g a FL. A d e s c r i p t i o n of this test is p r o v i d e d in t h e A p p e n d i x . The M e a n score of the 54 students w a s in the average range on the MLAT (M = 100.9, SD = 15.1). The students' scores r a n g e d from 72-126. PROCEDURE
The data on the predictor variables were collected at various intervals d u r i n g the c o m p l e t i o n of this study. The students were a d m i n i s t e r e d the WRMT-R, TWS, FRI, a n d P P V T - R at the foll o w i n g times: the b e g i n n i n g of first grade a n d at the e n d of the first, second, third, a n d fifth grades. The s t u d e n t s were administered the LAC at the b e g i n n i n g of the first grade a n d at the e n d of the first, second, a n d third grades. The L C O M P (listening c o m p r e h e n s i o n ) m e a s u r e w a s a d m i n i s t e r e d at the e n d of the third a n d fifth grades. The a f o r e m e n t i o n e d testing m e a s u r e s w e r e a d m i n i s t e r e d w i t h i n the first m o n t h a n d last m o n t h of each school year. All of the m e a s u r e s were a d m i n i s t e r e d individually to the participants by the first a n d third authors with assistance from g r a d u a t e a n d u n d e r g r a d u a t e s t u d e n t s trained b y t h e a u t h o r s for t h e p u r p o s e of this study. The s t u d e n t s " scores on the READINESS were obtained from school records at the e n d of their k i n d e r g a r t e n year. Students' scores on the TCS were obtained from school records at the e n d of the first grade, s The data on the o u t c o m e variables were collected at various intervals d u r i n g this study. The MLAT was a d m i n i s t e r e d by the
5. The authors extend special thanks to Mrs. Chris Utter and her staff for their very important roles in the completion of this study.
NATWE LANGUAGEPR~DICTOaS
141
first author at the beginning of the students' ninth grade year. The FL reading comprehension and FL writing measures were administered in groups by the first author at the end of the students' 10th grade year. The FL listening/speaking measure was administered individually by the fourth author of this study and graduate students trained by this author at the end of the students' 10th grade year.
RESULTS Individual analyses of variance (ANOVAs) were used to determine if there were significant differences among the three different FL groups (i.e., Spanish, French, and German) on the NL measures at each time period (first, second, third, and fifth grades), the MLAT, and the FL TOTAL PROF measure. Findings showed that there were no significant differences among the three groups on any of the predictor and outcome measures. Table I reports the scores of the participants on the predictor variables (i.e., NL and cognitive measures) used in this study. Because the purposes of this study were to determine the best NL predictors of FL proficiency and FL aptitude at five specific time periods, five separate analyses were conducted for each outcome variable. The time at which each analysis was conducted along with its corresponding label is PRE (beginning of first grade), POST 1 (end of first grade), POST 2 (end of second grade), POST 3 (end of third grade), and POST 4 (end of fifth grade). These labels (i.e., PRE-POST 4) for each of the five analyses are used throughout the remainder of this paper. Table II presents intercorrelations among the PRE test NL predictor variables and the FL outcome variables for the total group. The correlation patterns among the POST 1-4 predictor variables and the two outcome variables are similar in a general way and are not included here. A series of multiple regression analyses was applied to determine which of the NL and cognitive variables were important predictors of FL aptitude (MLAT) and FL proficiency (i.e., FL TOTAL PROF). The variable selection method, called MAXR, first determines the "best" one variable model (i.e., which single measure best correlates with a specific outcome variable). Then, the "best" two-variable model is selected on the basis of the maximum R2 (coefficient of determination) improvement. The "best" two variable model indicates the two measures that explain the highest percentage of the variance in FL proficiency
105.1
0.2
Test of Reading Readiness b
b Z scores
a Standard Scores, M = 100, SD = 15
0.1
Lindamood Auditory Conceptualization Test b
Test of Cognitive Skills a
104.7
102.1
16.2
86.6
Formal Reading Inventory a
Peabody Picture Vocabulary Test-Reviseda
Listening Comprehension a
97.0
1.8
1.0
14.0
14.1
0.1
104.9
104.6
11.4
79.3
14.3
91.8
Test of Written Spelling-2 a
M
Woodcock Reading Mastery Test-Revised a
SD
M
Testing Measures
Pre
Post I
1.0
13.8
17.3
13.4
14.2
SD
0.1
104.7
103.2
101.0
103.7
M
SD
1.0
13.4
18.5
11.9
13.4
Post 2
1.0
10.9
102.6
0.1
12.9
17.1
13.9
12.0
SD
108.4
107.9
101.2
102.1
M
Post 3
Table I. Means and Standard Deviations of ParticiPants on the NL Measures
13.9 109.7
105.2
12.5
13.5
13.4
110.5
11.8
100.8
SD
104.0
M
Post 4
p~
O3
bo
2. Test of Written Spelling
9. Foreign Language Proficiency Test * p < .05 ** p