Language Aptitude
NEW RESEARCH DIRECTIONS AND PERSPECTIVES
2. Understanding the meaning of passages: The six test items in this part
are identical in form to those in Section 1, but the assessment involves comprehension of whole passages rather than merely of lexical items. Again, half of the items are presented visually, the other half orally, and the passages differ in terms of the density of unknown words. The test differs from standard reading and oral comprehension tests in the inclu- sion of unknown words in the passages. Such words render these passages more like those that would be encountered in the process of learning an L2.
3. Continuous paired-associate learning: In this test, participants are pre-
sented with 60 paired associates (word pairs), half of them visually, half of them orally. They are required to learn the successive pairings and during this process they are tested at irregular intervals on words learned more recently as well as less recently. The test differs from a straight- forward paired-associates memory test in that there are certain rules that can facilitate learning, relating some of the terms to others.
4. Sentential inference: Participants receive 20 sets of three to five sen-
tences in the Ursulu language with their translations presented either visually or orally. They are then presented with a new sentence, either in English or in Ursulu, and are asked to indicate—based on inferences
tiple-choice answers best represents the translation.
5. Learning language rules: Participants are given some vocabulary, some
grammar, and some examples of how the Ursulu language works. From this type of information they are expected to learn some of the most evi- dent rules of the language. To measure this learning, they are presented with 12 items (lexical, semantic, morphological, and syntactic) that test their understanding of the Ursulu language.
Japan at exactly the same time, the Lunic Language Marathon, by Sick and Irie (2000), also focused on how the learners (seen as ‘imaginary space trav- elers’) master aspects of the hitherto unknown language of the planet ‘Luna.’ With regard to CANAL-FT, to validate the test, the authors report on a cor- relational study in which the convergent validity of the measurement pro- vided by the CANAL-FT was appraised by means of its correlations with the MLAT, and its discriminant validity was assessed through the test’s correla- tions with two established intelligence measures. The initial results were promising, indicating the viability of the CANAL theory; they also high- lighted future research directions, for example the need to look into how the test scores relate to working memory measures. In general, as the authors conclude, their “work should be viewed as a foundation for further develop- ment rather than as a completed effort” (Grigorenko et al., 2000, p. 401).
Sparks and Ganschow’s Linguistic Coding Differences Hypothesis
A systematic line of research by Richard Sparks, Leonore Ganschow, and their associates has focused on what they have labeled the Linguistic Coding Differences Hypothesis (LCDH). According to the hypothesis, one’s capac- ity to learn an L2 is closely related to the individual’s L1 learning skills, and L2 learning difficulties stem in part from native language difficulties (e.g., Sparks, 1995; Sparks & Ganschow, 1991, 1999, 2001; Sparks et al., 1995, 1998). The central cognitive factor the theory focused on is ‘linguistic cod-
ing,’ which refers to L1 literacy skills such as phonological/orthographic
processing and word recognition/decoding (i.e., single-word reading). The LCDH proposes that these abilities serve as the foundation for learning an L2, and an insufficient level of development in linguistic coding skills has a profound impact on L2 learning ability, resulting in a serious handicap. Thus, linguistic coding ability can be seen as a primary ID variable.
Sparks, Ganschow, and their colleagues have accumulated an impres- sive amount of evidence supporting their hypothesis and there is also some
data from other research that is in accordance with their conclusions. One of the main types of LCDH studies has involved conducting comparative analyses of good and poor L2 learners in various age groups and learning situations to see whether they differed in their linguistic coding skills. As Sparks and Ganschow (2001) summarized, the findings consistently revealed that (a) successful L2 learners exhibited significantly stronger L1 literacy skills than unsuccessful learners, and they were also superior on L1 syntactic measures but not on semantic tasks; and (b) successful L2 learners had sig- nificantly stronger language aptitude (measured by the MLAT). In one study the researchers also found that L2 word recognition was one of the best pre- dictors of English proficiency. Based on these findings Sparks and Gan- schow recommend that one important way of improving language aptitude measures is to elaborate on phonological measures both in the learners’ L1 and L2 by including in the instruments relevant tasks such as word recogni- tion, pseudoword decoding, phonological memory, and phonemic aware- ness.
The significance of L1 literacy skills in L2 studies has also been high- lighted in an important longitudinal study conducted by Dufva and Voeten (1999), investigating 160 Finnish elementary school children from the first to the third grade. The researchers examined two cognitive areas, L1 literacy acquisition and phonological memory (the latter being part of ‘working memory’ and is discussed next) in terms of their impact on learning English as a foreign language. The longitudinal design of their investigation allowed them to establish cause–effect relationships, which they then confirmed us- ing structural equation modeling. Measures obtained in Grade 1 were L1 word recognition and listening comprehension; in the second grade L1 word recognition, reading comprehension, and phonological memory; and in the third grade, L2 skills. The authors found that both L1 literacy and phono- logical memory had positive effects on L2 learning, together explaining 58% of the variance in English proficiency. This is a remarkably high figure (cor- responding to a multiple correlation of around 0.76), rarely encountered in L2 studies, and its magnitude is particularly noteworthy because the tasks measuring English proficiency covered a wide range of competencies, fo- cusing on listening comprehension, communicative skills, and active vo- cabulary knowledge.
The authors have also found that it is not the rate of development in word recognition from first to second grade that mattered but the ultimate
level achieved by the end of Grade 2: The faster the children’s speed of word
recognition was, the better they were at English. In fact, the level of development of L1 word recognition was the strongest predictor of L2 learning in the whole study. Based on these results, Dufva and Voeten (1999) concluded that native language word recognition formed the basis of learning an L2. Therefore, in agreement with Sparks and Ganschow (2001),
they recommended that educators assess L1 literacy skills early on so that at- risk children can be provided with intensive literacy instruction, especially in the prerequisites of word recognition (e.g., phonological awareness), in order to enhance their learning an L2.
An additional element of Sparks and Ganschow’s (1995, 1999; Sparks, Ganschow & Javorsky, 2000) overall thesis has been that affective differ- ences between students with lower and higher levels of L2 skills (e.g., dif- ferences in anxiety) are a consequence of their differing levels of self-per- ceptions about their L2 learning skills. Thus, as they claimed, students with high levels of anxiety, poor language attitudes, and low motivation in the L2 classroom are likely to be those who have a history of and current difficulty with some aspect of their native language skills. This view has been strongly contested by MacIntyre (1995a, 1995b, 1999) and Horwitz (2000; for more detail, see the discussion on anxiety in chapt. 7).
Working Memory and Language Aptitude
Research into the relationship between working memory and SLA appears to be one of the most promising current directions in language aptitude studies, and as Miyaki and Friedman (1998, p. 339) conclude, “working memory for language may be one (if not the) central component of this language apti- tude.” Memory has traditionally been part of every aptitude model but with a complex domain such as memory the secret lies in the details, that is, how the construct is conceptualized. With regard to the memory component of the MLAT, even Carroll (1990) admitted that he had never been completely confident about the validity of the rote-learning subtest. And given that memory research has made enormous progress during the past two decades in cognitive psychology, we can safely conclude that this area lends itself to improvement in modern aptitude testing.
Working memory involves the “temporary storage and manipulation of
information that is assumed to be necessary for a wide range of complex cognitive activities” (Baddeley, 2003, p. 189); thus, it underpins our capacity for thinking and has important specific implications for language processing. Gathercole and Thorn (1998), for example, drew attention to the possible contribution of the verbal component of working memory, the ‘phonological loop,’ to the learning of the sound patterns of new L2 words. Indeed, the concept of working memory appears to be an ideally suited memory construct for SLA purposes because besides its phonological short- term memory constituent it also comprises a featured ‘attention’ component, and the role of attention and attentional capacity has been a key research target in recent SLA research (cf. Robinson, 2003). For this reason, Ellis
(2001) emphasized that the concept deserves much more consideration than it has been thus far given in L2 studies.
In an analysis of the role of working memory within the language apti- tude complex, Sawyer and Ranta (2001) agreed that working memory ca- pacity may be the key to elaborating the concept of language aptitude and to integrating aptitude into the SLA process. They argue that the concept has several advantages over the traditional short-term memory construct, which it has replaced in the literature, because working memory is more than a pas- sive transitional storage stage preceding long-term memory: It is a more ac- tive system, comprising an independent temporary cognitive workspace or ‘computational arena’ (Miyake & Friedman, 1998) with ongoing processing functions (hence the ‘working’ part of the term), used for sequential cogni- tive processes, such as the comprehension and production of language.
How is the construct of ‘working memory’ structured? In 1974 Baddeley and Hitch proposed that it could be divided into four subsystems (for a recent review, see Baddeley, 2003):
(1) The phonological loop is the specialized verbal component of working memory, concerned with the temporary storage of verbal and acoustic information. The stored material is subject to rapid decay (over approxi- mately two seconds) but the loss of information can be offset by ‘subvo- cal rehearsal,’ which reactivates the decaying representations and which can also translate visual information into phonological form.
(2) The visuospatial sketchpad is the visual equivalent of the phonological loop, responsible for integrating spatial, visual, and kinesthetic informa- tion into a unified representation, which can be temporarily stored and manipulated. This system is involved, for example, in everyday reading tasks but its functioning has been less studied than that of the phonological loop. Baddeley suggests, however, that similarly to the phonological loop, the visuospatial sketchpad also has a storage and a processing component (the letter termed the ‘inner scribe’) which can, for example, translate verbal information into an image-based code. (3) The central executive is the most important and least understood aspect
of working memory, responsible for its attentional control. It constitutes the supervisory attentional system that allocates attentional resources and regulates the selection, initiation, and termination of processing routines (e.g., encoding, storing, and retrieving). Thus, it receives, coor- dinates, and integrates information from the subsystems of the visu- ospatial sketchpad and the phonological loop as well as from long-term memory to carry out complex cognitive tasks such as future planning, decision making, mathematical calculations, and reasoning. It is also in- volved in performing reading and comprehension, and—interestingly—
in trouble-shooting in situations in which the automatic processes run into difficulty, which links it to the use of communication strategies (cf. Dörnyei & Kormos, 1998). The central executive is of particular rele- vance to our discussion because the executive processes are thought to be the principal factors determining individual differences in ‘working memory span’ (Baddeley, 2003; Daneman & Carpenter, 1980).
(4) The episodic buffer has recently been added to the working memory construct to constitute a storage counterpart of the central executive, which is now seen as a purely control system without any storage ca- pacity. The episodic buffer combines information from different sources and modalities into a single, multi-faceted code, or ‘episode’—hence the ‘episodic’ part of the label. It is assumed to underpin the capacity for conscious awareness.
The overall capacity of working memory can be expressed in terms of the working memory span. This has proved to be a robust predictor of a wide range of complex cognitive skills and it is highly correlated with perform- ance on the type of reasoning tasks that underpin standard tests of intelli- gence. It is measured by instruments and procedures whereby participants are typically required to combine some sort of (a) processing and (b) storage of information in a dynamic and simultaneous manner; thus, the assessment goes beyond traditional memory tests such as digit or word span measures. Table 3.4 describes one of the best-known instruments developed by Dane- man and Carpenter (1980).
Miyake and Friedman (1998) emphasized that although working mem- ory plays a central role in all forms of higher-level cognition, its role is par- ticularly featured in language processing because both the production and the comprehension of language requires the processing of sequences of sym- bols over time in a linear manner. This linearity inherently necessitates a temporal storing capacity and the ability to integrate information from the stream of successive discourse. According to the current conceptualization, working memory matches these simultaneous processing and storage re- quirements perfectly. The authors therefore conclude that individual differ- ences in LI working memory capacity for language are closely related not only to L2 working memory capacity and L2 language comprehension skills but also to the speed and efficiency of the acquisition of L2 knowledge.
Table 3.4. Description of Daneman and Carpenter’s (1980) Reading Span Test
Participants are asked to read aloud a set of unrelated sentences and then recall the final word of each sentence in that set. The 9 to 16 word long sentences were taken from general knowledge quiz books and each ended in a different word; e.g.,
• “You can trace the languages English and German back to the same
roots.”
• “The Supreme Court of the United States has eleven justices.”
The processing element of the test is provided by the task that after reading each sentence the participants have to decide whether it was true or false—the sentences are of moderate difficulty, with half of them being true and the other half false.
The total test contains three sets each of two, three, four, five, and six sentences and the participants are presented increasingly longer sets until they fail to recall the sentence-final words of all three sets at a particular level.
The level at which a participant is correct on two out of three sets is taken as a measure of the individual’s reading span. Being correct on only one set at a particular level is given a credit of .5. Miyake and Friedman (1998) added that in some studies the reading span measure has been the total number of sentence-final words recalled from all the trials.
The test also has a listening version, which works along the same lines and which correlates well with the reading span.
Sawyer and Ranta (2001) reported a number of studies that have demonstrated a fairly strong relationship between working memory capacity and L2 proficiency, as well as between L1 and L2 working memory capac- ity. Their conclusion is quite positive:
If further research confirms WM [working memory] capacity to be a stable individual difference among learners, its potential importance in SLA will be clear. Assuming that noticing is crucial to learning, and at- tention is required for noticing, and attention at any moment is limited by WM capacity, then there must logically be a close relationship be- tween amount of learning and size of WM. It is also likely that WM serves as an arena in which the effects of other components of aptitude are integrated. (p. 342)
Finally, let me return to Segalowitz’s (1997) synthesis of SLA, cogni- tive, and neuropsychological frameworks mentioned earlier. Based on his
broad review of the literature, Segalowitz highlights the paramount impor- tance of neurocognitive flexibility for the successful handling of the com- plexity of linguistic input. As he argues, the full processing of linguistic in- put requires the recognition of a number of diverse aspects, ranging from sound sequences through syntactic patterns to nonlinguistic cues, and it also requires the coordination of one’s own communicative responses in a situa- tionally appropriate manner. Thus, Segalowitz continues, effective L2 per- formance entails rapid shifting of attention from one communicative dimen- sion to another, which presupposes a “high degree of perceptual and produc- tion fluency and the requisite neurocognitive flexibility to accomplish this” (p. 104). It seems to me that the degree of flexible attentional and resource control that Segalowitz outlines is to a large extent the function of one’s working memory capacity, which again underlines the importance of the concept of working memory regarding second language acquisition and use.
Robinson’s Research on the Aptitude–Treatment Interaction
A central issue in ID research, and one that has emerged in aptitude research in particular, is the question as to whether there are any optimal combina- tions of ID variables that are especially conducive to efficient learning. Of course, this question would not have risen if some scholars had not assumed that such constellations of learner traits existed; one researcher in particular, Richard Snow, was influential in highlighting the potential importance of such ID variable clusters, or as he called them, aptitude complexes. His ini- tiative has been taken up by several of his colleagues and students (cf. Ac- kerman, 2003; Corno et al., 2002) because, “Although isolated traits often have … substantial impact on learning outcomes, it may be that combina- tions of traits have more predictive power than traits in isolation” (Acker- man, 2003, p. 92). Furthermore, the concept of ‘aptitude complexes’ can also be combined with Cronbach’s ‘aptitude–treatment interaction’ approach that concerns the ways by which mental abilities interact with learning con- ditions, resulting in a powerful situated ID paradigm for learning. This is the theoretical foundation that Peter Robinson (e.g., 2001, 2002d, in press) drew on when he launched his pioneering research program on language aptitude- treatment interaction. He conceptualized language aptitude as the sum of lower level abilities, grouped into cognitive factors, which differentially support learning in various learning situations/conditions.
We have already seen that the relationship between language aptitude and learning methods/situations has traditionally been a focal issue in apti- tude studies; indeed, in his review of the history of language aptitude re- search Spolsky (1995) pointed out that scholars claimed already in the 1930s that language aptitude could only be defined in the context of the teaching
method that was to be used. Segalowitz (1997) approached the question from a very different perspective—the synthesis of SLA, cognitive, and neuropsychological frameworks—yet arrived at the same conclusion: “what we perceive as language learning ability is not a fixed characteristic of a per- son but rather a complex reflection of the whole learning situation” (p. 108).