Second language learning: extent of acquisition and retention

¶ … Second Language Learning

To What Extent May L1 Affect Second Language Learning

Linguistic and Metalinguistic Knowledge

This category includes variables that are effective in both reading and listening comprehension and that involve knowledge about the structure of language, such as its syntax and morphology. Two questions guide the discussion here: How does linguistic knowledge in L2 develop, and how does linguistic knowledge in L1 affect L2 linguistic knowledge, indicating cross-language transfer?

Syntactic Knowledge. The development of syntactic knowledge has been one of the most productive research areas in applied linguistics, especially in the field of second language acquisition. A typical study involves selecting a linguistic dimension (for example, relative clause formation strategies) and then comparing groups of bilinguals who have different ways of representing that parameter in their L1 (Robert & Williams, 2009). These studies tend to emphasize the Universal Grammar underlying all languages and suggest that second language acquisition involves setting new values for the universal parameters. Initially, L1 parameters are used in interpreting L2 but, with experience, new values are set for L2. For example, Jackson, (1981) focused on the adjacency conditions in English and French. In English, the adjacency requirement is much stricter than in French. That is, a verb and its direct object need to be next to each other. In White's study, English and French speakers made grammaticality judgments to French and English sentences that did or did not violate the adjacency assumption. French speakers judged adjacency requirements more flexibly, even with English sentences, whereas the English speakers judged the adjacency requirement more strictly for both French and English sentences. In short, L1 strategies were used to interpret L2 sentences. However, some researchers suggest that L2 learning is different from L1 acquisition, and that Universal Grammar plays a minimal role in L2 learning, especially for adults (Dodson, 1985). Also note the distinction made between acquisition vs. learning in this discussion, where the former, but not the latter, implies innate, nondeliberate processes). A defining moment for the field was in the late 70s / early 80s when it became evident that L2 learners follow a fairly rigid developmental route, in the same way as children learning their L1 do, and not dissimilar in many respects from the L1 route (Myles, 2002).

The competition model (Anton & DiCamilla, 1998) has been the most systematic way of describing how the specific L1 syntactic strategies affect sentence comprehension in L2. This model, as discussed in Jackson, (1981), de-emphasizes linguistic universals and focuses on the cues in each language that are the most salient in assigning syntactic roles to words. Although the model is developed through listening comprehension studies, it can also apply to reading comprehension. (With both child and adult monolingual readers, listening comprehension explains a large amount of variance in reading comprehension; e.g., Cook, 2001).

In the basic experimental tests of the competition model, participants listen to sentences created by different orderings of two nouns (N1 and N2) and a verb (N1VN2, N1N2V, VN1N2, and so on). Of course, some of these orderings may be ungrammatical in the language that is being tested. The subjects then try to identify the agent of each sentence. In English, word order is a very salient cue, and, usually, the noun that comes before the verb is assigned the subject role. If the English sentence The scarecrow bit the dog is read or heard, the role of the subject doing the biting is assigned to the scarecrow, although in real life this is an unlikely event. In other languages, other cues are more salient (e.g., animacy in Italian and German, inflections/particles in Turkish and Japanese). The interesting point is that, when a bilingual is trying to figure out the meaning of a sentence, two sets of cues (those from L1 and L2) can compete in assigning the syntactic role. This competition can highlight the nature of linguistic knowledge that guides syntactic processing (Robert & Williams, 2009).

In several studies by McDonald ( 1987, 1989), Dutch learners of English and American learners of Dutch were observed. In general, novices used L1 cues to interpret L2 sentences, but, as they became more proficient, they began to use L2 cues just like native speakers of that language. However, this pattern is not always clear.

A study by Cook, (2001) with adult English speakers learning Japanese and Japanese speakers learning English, indicated that cross-language transfer of cues is asymmetric. English speakers began to use Japanese cues more quickly and approached the general pattern of Japanese native speakers. On the other hand, Japanese speakers analyzing English sentences were slower in adopting the salient English cue of word order. The influence of the Japanese flexible word order was still evident. Based on these data, Sasaki argued that lexical-semantic cues (such as animacy) are more salient than syntactic cues, such as word order, and that these salient semantically-based cues in L1 are harder to drop when analyzing L2 sentences.

However, Cook, (2001) had used uninflected sentences, and these are unnatural in Japanese. This may have caused the animacy cues to become artificially more salient in the absence of case-marking particles. Corder, (1998) asked American, Chinese, and Korean college students who were learning Japanese to decide on the actor of sentences. Koda's sentences varied on two dimensions: presence of case-markings and whether the word order was canonical (SOV for Japanese). Across all language groups, case-markings improved sentence comprehension. However, word order affected Chinese and American groups (whose languages emphasize strict word order) more than the Korean group. Korean participants performed equally well on canonical and noncanonical sentences, but the non-Korean participants did better on canonical sentences. This study also showed the link between syntactic analysis and reading comprehension because, for all subjects, reading comprehension was strongly related to the knowledge of particle markings.

Finally, an interesting twist was reported by Choong, (2006) who discovered that English-speaking students of Japanese learn very quickly that the first noun may not be the subject in Japanese. However, on NNV sentences, they begin to use a much stricter subject-object-verb rule (when compared to native Japanese speakers), because they still treat word order as a crucial dimension but they change its values, parallel to the parameter setting view discussed earlier. To summarize, syntactic knowledge can affect reading comprehension in an intricate way. Both L1 and L2 grammatical cues may play a role in how syntactic roles are assigned when reading texts in either language.

Metalinguistic Awareness and Phonological Awareness. Syntactic knowledge requires a reader to become aware of the systematicities in the language and the formation of structures. In addition to an awareness of syntactic structures, awareness of other characteristics of language is needed. When we use language, we use it as a transparent tool to communicate some meaningful message, hardly paying attention to the form of the message. Metalinguistic awareness is "the ability to make language forms opaque and attend to them in and for themselves . . . one which makes special cognitive demands and seems to be less easily and less universally acquired than the language performances of speaking and listening" ( Cazden, 1992, p. 62). In other words, metalinguistic awareness requires attending to the structural characteristics of a language rather than just using the language. Such an analytic focus on language plays a major role in literacy acquisition, especially for young children.

Researchers have discovered that bilingual children show a better awareness of the distinction between form and meaning than monolingual children, for example, in realizing that words are actually arbitrary labels and are not tied to an object (Cook, 2002). Even more impressive are the gains in kindergarten children's metalinguistic awareness assessed after foreign language instruction that lasts as little as 20 minutes per week for several months (Cook, 2001)!

When it comes to reading, monolingual research has convincingly demonstrated the importance of phonological awareness in the development of a child's reading skills (Dodson, 1967). This type of metalinguistic awareness is the ability to hear the subsegments, such as phonemes and syllables, in the spoken language. If young children can notice the subcomponents of a spoken language, they will have less difficulty in mapping letters to these speech segments when they are learning to read. For bilingual children, how this awareness develops in both languages is the question of interest. Some of the research in this area is specifically concerned with the transfer of phonological awareness. Those studies are discussed in the next section.

Vocabulary and Morphological Analysis. Although second language acquisition research focused more on syntax than on vocabulary, some of the reading difficulties for second language readers may stem from limitations in vocabulary rather than in syntactic knowledge (Dodson, 1985). In fact, how they determine the meaning of unknown vocabulary in Spanish and English was one of the key differences between a proficient and a nonproficient bilingual reader in a study by Dodson, (1985).

When (Choong, 2006) analyzed the scores of 270,000 students (across a period of 10 years) on the English reading and listening comprehension tests given nationally in Finland, he noticed an interesting pattern. On the same English tests, Swedish-speaking Finns consistently did better than Finnish speakers. Because the students came from similar backgrounds, one likely explanation was that, when completing the English tasks, the Swedish speakers get more help from their L1 than Finnish speakers, as Swedish, but not Finnish, is related to English.

Supporting this view is Robert & Williams, (2009) finding that Swedish children already knew a lot of English words when they started instruction in English because of the cognates in Swedish and English. However, it is also possible that some readers are not aware of cognates and cannot exploit this source of information to facilitate their reading comprehension (Cook, 2001). In a recent study, we focused on the English reading comprehension of upper elementary Spanish-English bilingual students (Choong, 2006). The question was whether knowing the Spanish cognates of some key words would help in comprehending the passages in English. Results indicated that just knowing the Spanish word was not enough; an awareness of its cognate status was also necessary before this knowledge could help English comprehension.

When they are faced with an unknown word in L2 text, an understanding of the morphological structure of that language can help bilingual readers discover the meaning of the unknown word. Therefore, knowledge of L2 morphological structure can help in vocabulary development. Dodson, (1967) investigated the morphological knowledge of Spanish-English bilingual children in Grades 4, 6, and 8. Vocabulary knowledge (as measured by a yes/no test on a list of words) was strongly correlated with morphological knowledge. Morphological knowledge was assessed in a translation task to determine if the children were aware of the systematic relationship between English and Spanish suffixes (felicity-felicidad). Performance was higher on cognates (e.g., felicity-felicidad) than on noncognates (e.g., dangerously-peligrosamente). In addition, the ability to analyze the suffixes on both cognates and noncognates improved considerably with grade, indicating a developmental trend.

Because vocabulary knowledge develops with experience, the type of exposure to L2 becomes an important factor. As Jackson, (1981) discussed, contexts in which a word's meaning is learned may affect what is learned. For example, formal instruction may highlight different aspects of a word's meaning as compared to informal personal interactions. Production activities that follow these NS models of interaction should be task-based, in that they should mirror real-world activities in which NSs are often engaged (Doughty & Long 2003). A dramatic difference in vocabulary acquisition is observed when children who acquire L2 from their surroundings are compared with adults learning L2 in a formal classroom. Why children acquire vocabulary with such ease has been debated for a long time. One of the explanations given, among others, is the input hypothesis (Corder, 1998). According to the input hypothesis, the nature of the input that the children receive when they are acquiring their L1 is contextualized here and now and limited in scope of topics. Such contextualization makes the input more comprehensible (Robert & Williams, 2009) and easier to learn. Hence, the issue of comprehensible input has been raised in designing formal L2 programs for adult L2 learners. Current popularity of programs that emphasize communication rather than grammatical knowledge is a case in point.

Another issue in vocabulary knowledge as it relates to reading comprehension is the difference in the structure of word meanings across languages (Choong, 2006). For example, the core meaning as well as the associations for the French word rendezvous cannot be fully captured by the English words date or appointment. To distinguish between such shades of meaning and, thus, to comprehend the nuances of the text, one also needs some background knowledge and familiarity with the culture. This topic is discussed further in the Prior Knowledge subsection.

Literacy Variables

Although listening and reading comprehension are closely linked because of the common linguistic base, there are several major differences between oral and written language (for a review, see Olson & Torrance, 1991) that merit a separate discussion of variables related to the comprehension of written language. Written language is decontextualized, that is, contextual cues, such as the paralinguistic ones found in face-to-face oral communication and immediate feedback if breakdown in comprehension occurs, are missing (Jacobson, 1990).

One measure of decontextualized language skills is the ability to provide formal definitions. Being able to use language in a decontextualized manner, such as in providing formal definitions, is a strong predictor of reading comprehension for bilingual children (Cook, 2001). Ironically, in contrast to the current emphasis on contextualization and communicative competence, formal vocabulary knowledge requiring decontextualized language use is more strongly related to L2 reading proficiency than oral proficiency is. Such a reification of language does seem to discriminate both among native speakers and among non-native speakers; it does, of course, have some bearing on our first message, that of skill, since there may well be an element of knowledge within skill which determines differential proficiency (Davies 2006).

Also included under literacy variables are dimensions globally described as good meaning-making strategies (Cook, 2001) or processes of negotiating with text (Prodromou, 2003). These strategies extend beyond the linguistic variables, discussed in the previous section, that could apply to both written and spoken language.

In this category are variables that are unique to reading, such as knowledge about the uses, purposes, and conventions of literacy; and the ability to monitor comprehension while reading, and to exploit text structure and signaling devices. Because reading is an active process, readers have to continuously watch for information that they need, detect inconsistencies in their representation and interpretation of the text, and take steps to correct any problems that occur.

Cook, (2002) included the following dimensions as good meaning-making strategies: (a) envisionment building (creating a representation of the text and updating it as more is read), (b) hypothesizing (e.g., predicting what comes next), (c) understanding the text's language (L1 or L2), and (d) being familiar with genre characteristics (e.g., a report vs. A story). In a qualitative study, they observed 12 fifth-grade bilingual students as they read texts of different genres and languages. The researchers discovered that, after some basic linguistic knowledge was in place, the meaning-making strategies made the most difference in how well each reader comprehended and remembered texts. The good meaning-making strategies were also related to how the students performed on standardized reading tests. Interestingly, good meaning-making strategies transferred across languages, indicating that, once the strategies are in place, they can be used to comprehend texts in both languages. This implies that, for adult L2 learners already literate in their L1, good meaning-making strategies may already be in place, and some of these literacy variables may carry less weight.

In their case study comparing three children, two bilingual readers (one proficient and the other not proficient) and one monolingual (proficient) reader, Jimenez et al. ( 1994) discovered that, when discussing aloud a text that was just read, both of the proficient readers used many strategies that were missing from the think-aloud protocols of the less proficient reader. Proficient readers emphasized reading for comprehension, and constantly monitored their comprehension. In addition, these readers persevered when faced with noncomprehension. The less proficient reader, on the other hand, wanted to keep on reading and reach the end of the passage regardless of whether the passage was comprehended. The proficient readers reported using strategies that included rereading, visualizing, and invoking prior knowledge. More interestingly, the proficient bilingual reader had some strategies that differed from those of the monolingual proficient reader. The bilingual reader emphasized the importance of vocabulary knowledge much more and spent more time on figuring out word meanings. In addition, this child used L1 (Spanish) to comprehend L2 (English) texts, especially by exploiting cognates (Jacobson, 1990).

Good meaning-making strategies were observed in several adult reading comprehension studies as well. In a study with adult learners of English (with varying first languages), Choong, (2006) were interested in the teaching of good meaning-making strategies. The researchers trained two groups of participants on different metacognitive strategies. One group of students learned about semantic mapping, that is, activating and graphically organizing ideas about a topic before reading the text on that topic. Another group of students had ETR (experience-text relationship) training. These students discussed their thoughts on, and experiences with, a topic before actually reading the relevant text. Finally, a control group had no metacognitive training. All participants read the same texts and had pre- and post-tests on the texts. The three groups showed similar gains between pre- and post-test performances on multiple choice comprehension questions. On open-ended questions, only the two training groups showed significant gains.

For skilled readers, another source of information in the written text is its structure. For example, stories tend to follow a certain outline, as discussed in story grammar research. Awareness of a text's structure can help readers organize their mental representation of the text (Prodromou, 2003).

Corder, (1998) observed 120 English-speaking college students taking various levels of French courses. All subjects read French texts in structured (i.e., following story grammar) or less structured (episodes with flashbacks interspersed or episodes completely mixed, as in a modern short story) formats. Third-year students had equally high levels of recall for all three text types. However, first-year students recalled the most information from the structured text and the least from the completely mixed-episodes text. That is, text structure interacted with a reader's proficiency level. In addition, Prodromou, (2003) discovered that training on text structure may not work equally well for all types of texts. Taken together, these studies indicate that text structure may not be a strong factor by itself but may depend on reader and text characteristics. The reader variables, especially the existing knowledge structures of the reader, are the focus in the next section.

Evolution of theories of second language acquisition

One of the maxims from monolingual reading research is that how much one already knows about a topic affects how much one will comprehend a text on that topic. This presents a special challenge for bilingual readers because, when reading in their L2, these readers may not have the background knowledge, assumptions, and even stereotypes taken for granted by the writer of the text. Hence, the L2 reader-writer interaction may be quite different from the L1 reader-writer interaction. If L2 readers (both child and adult) have the general cultural framework assumed by the writer, then they can easily comprehend a text and make the necessary inferences. How background and world knowledge affect L2 reading has been studied extensively (Cook, 2002). In a review, Bernhardt ( 1991b) classified this research into three categories. In the first group of studies, cultural background is used as the quasi-independent variable, and how readers comprehend texts from their own and others' cultures is compared. In a classic study of this type, Dedrinou, (2006) gave English texts, describing either an American wedding or an Indian wedding, to American or Indian adults. Comprehension was higher when the text matched the cultural background of the reader. More importantly, on mismatched texts, readers made incorrect inferences and distortions.

In the second group of studies, readers' background knowledge is assessed or assumed from schooling (such as field of study for college students), followed by the reading comprehension test (Brooks & Donato, 1994). Jacobson, (1990) assessed the prior knowledge of upper elementary Spanish-English bilingual and English monolingual students before giving them an English reading comprehension test. The data showed that prior knowledge of the topic was significantly related to reading comprehension performance. If the effects of prior knowledge were partialled out, then reading comprehension scores did not vary among the groups of monolingual and bilingual students.

In the third group of studies, the background knowledge is defined more narrowly. In these studies, ambiguous passages are given, with or without disambiguating information (such as a title or a picture). For example, Prodromou, (2003) found that providing background information to adult students of English-as-a-Second-Language (ESL did not affect recall of the passages in English. Choong, (2006) replicated the same study but asked the participants to recall the texts in their L1. This time, the effects of disambiguating contexts were observed. Hence, Lee suggested that recall in L1 of the readers may more accurately portray the influence of background knowledge on reading comprehension in L2.

Although the role of background knowledge on reading comprehension is well researched, several caveats are in order in terms of L2 readers. Lado, (2004) emphasized that culture-specific knowledge contains cultural historic, ritualistic knowledge that is transmitted within a culture. It may be more difficult to isolate and identify this knowledge source compared to domain-specific knowledge, such as technical knowledge about the American government.

A possible problem exists if the presence or absence of background knowledge is judged on the ethnic/cultural backgrounds of participants. For example, if a study is comparing how Muslims and Catholics comprehend texts on specific religious topics, it ignores the fact that people may be familiar with faiths other than their own (Dedrinou, 2006).

In addition, Lado, (2004) demonstrated that, even when familiarity is assessed by a test rather than reliance on ethnic/cultural backgrounds of participants, familiarity with a topic may not be determined easily. In Ham madou's study, English-speaking college students were learning French or Italian as a foreign language. Before reading the three target passages, they rank-ordered the topics of the passages on familiarity. The recall of the passages was affected by the proficiency level of the subjects (as determined by which course in a sequence they were taking), but not by topic familiarity. One can assume that rank ordering was not a sensitive measure of familiarity. However, the same pattern was found by Lado, (2004) even though she had measured the familiarity more thoroughly by free associations to a topic.

OTHER ISSUES

These three components -- linguistics, literacy, and prior knowledge -- play a role in both monolingual and bilingual reading comprehension. However, in L2 reading, these factors are doubled. Both L1 and L2 variables play a role in L2 reading. In addition, sociocultural forces help shape L2 literacy acquisition and development. This section focuses on issues that are unique to L2 reading and literacy.

Different Fields Studying Bilingualism or Tyranny of Participant Populations

One of the most interesting things one notices while reviewing literature on bilingual reading is the wide gap between the two fields that study the same topic. In applied linguistics, there is mushrooming research on ESL and foreign language readers. Likewise, bilingual education literature is full of research on language-minority readers. However, these two groups rarely talk to each other. Research housed under applied linguistics has a pretty strong top-down flavor, emphasizing variables, such as background, discourse and text structure, and downplaying the role of lower level processes, especially word recognition. Research from the bilingual education tradition, on the other hand, has a strong sociocultural flavor, emphasizing the social and political contexts affecting bilingual reading.

Of course, part of this disconnectedness stems from the different participant populations in the two fields. In applied linguistics, the focus is usually on college students who are already literate in their L1, are highly educated, and have chosen to become literate in L2. In bilingual education studies, the participant populations are usually young children who speak a minority language and did not have a choice in whether to become literate in their L2. These children may or may not be literate in their L1 (Corder, 1998). In between these two extremes are new immigrants who may or may not be literate in their L1. (ESL studies sometimes focus on these immigrants, although most ESL research is done with foreign students learning English to continue their higher education). In this paper, both research traditions have been included because, as long as the reader characteristics are identified as a quasi-variable (or as the contextual condition), we can learn a lot about the cognitive processes in bilingual reading when we integrate the findings from different fields with different types of bilinguals.

The Transfer Question

Several different studies have indicated that the level of academic achievement in the first language (most likely to be caused by formal instruction in the first language) is correlated with increased academic proficiency in the second language, even when the two languages have different writing systems (Lado, 2004), but, if there are problems in L2 reading comprehension, what is the cause -- limited L1 literacy or limited L2 linguistic knowledge? In a seminal paper, Prodromou, (2003) focused on the reasons for poor L2 reading comprehension and asked whether it is a reading problem or a language problem (Dedrinou, 2006). One possibility is that individuals who read poorly in L2 may do so because either they do not have good reading skills in L1 or they cannot transfer them to L2. This view assumes reading to be a universal process across languages. On the other hand, a certain threshold of linguistic proficiency may have to be established before L1 reading comprehension skills are transferred to L2 (Brooks & Donato, 1994). These two hypotheses have different educational implications ( Alderson). If the source of the problem is assumed to be poor linguistic knowledge, then instruction would focus on developing linguistic competence. If, however, the source is assumed to be poor reading strategies in L1 (and through it, in L2), then the instruction would focus on appropriate reading strategies.

Prodromou, (2003) in a study with Turkish students in the Netherlands, assessed reading comprehension of these participants in Dutch and Turkish as well as their knowledge of Dutch syntax and vocabulary. The subjects were all graduates of Turkish high schools, and they were highly literate in their L1. Bossers determined the predictors of L2 reading comprehension using multiple regression analyses. The L1 reading comprehension and L2 linguistic knowledge were the variables in the equation. Overall, both L1 reading proficiency and L2 linguistic knowledge were significant predictors. However, the predictive power of L2 linguistic knowledge was several times stronger. In a post hoc analysis, participants were classified as skilled and less skilled readers of Dutch, and the same two variables were used to predict L2 reading performance. For less skilled readers, L2 linguistic knowledge was the main predictor of L2 reading comprehension. On the other hand, for skilled readers of Dutch, L1 reading comprehension was the significant predictor of L2 reading (Lado, 2004).

These studies provide us with the initial data on the link between linguistic and literacy knowledge. However, to address the transfer question more thoroughly, the relationship between language and literacy development in the two languages of a bilingual should be studied at a more detailed level of analysis, especially focusing on what transfers from L1 to L2 reading.

The issue of transfer has even stronger educational implications for young readers. For example, in bilingual education programs, students who do not have the majority language as their L1 acquire literacy in their L1 while learning L2. When they are assumed to have enough L2 proficiency (more on this issue later), they are transitioned to all-L2 classrooms (Dedrinou, 2006). One of the basic assumptions underlying bilingual education is that literacy development in L1 of the child will progress more rapidly because of support from the family, culture, and language. This assumption has been confirmed. Another basic assumption of bilingual education is that the development of literacy skills and knowledge in one language will transfer to the other language. In some global studies of this question, a significant correlation has been shown between reading proficiency in L1 and L2 (Lado, 2004).

In the last couple of years, my colleagues and I have been focusing on the specific nature of transfer. We have been investigating different components of the reading process to determine how L1 variables affect L2 outcomes. In this area, likely candidates are metalinguistic and metacognitive awareness, components of reading that are relatively language-independent and can transfer across languages easily. As discussed earlier, one of the strongest predictors of literacy acquisition is phonological awareness. Children who can hear the small components in the spoken language find it much easier to map the written symbols to the spoken language and, hence, learn to decode. This pattern has been replicated many times in English and in different alphabetic languages with monolingual readers (Corder, 1998). Our question was whether the developing phonological awareness in the L1 of a bilingual beginning reader will transfer and affect word recognition in the L2 of the child (Lado, 2004). The results showed that children who had high levels of phonological awareness in Spanish were more likely to decode English words and pseudowords easily. In other studies, we identified significant correlations between phonological awareness in the two languages of bilingual children (Prodromou, 2003).