Presson and MacWhinney - Second Language Grammar

From LearnLab
Revision as of 22:27, 4 December 2009 by Macw (talk | contribs) (social disincentives)
Jump to: navigation, search

Spanish Conjugation

Summary Table


Adult second language learners often fail to acquire enough fluency in the new language to support smooth communicative interactions. The studies described here explore the hypothesis that robustness can be markedly improved through basic skill training based on three related pedagogical methods: graduated interval recall, resonant co-training, and cue focusing. This prediction will be tested in the context of in vivo and laboratory studies of online learning of Spanish verb conjugation.

Background & Significance

The central controversy in the study of second language acquisition is the status of the Critical Period Hypothesis. As formulated first by Penfield & Roberts (Penfield & Roberts, 1959) and then later by Lenneberg (1967), this hypothesis holds that, after some critical age, second languages (L2s) cannot be learned to full native-speaker competence. This critical period has been variously linked to age 2 for lexical learning (Weber-Fox & Neville, 1996) and perception (Kuhl, Conboy, Padden, Nelson, & Pruitt, 2005), age 6 for phonology (Flege, Yeni-Komshian, & Liu, 1999), age 13 for syntax (Johnson & Newport, 1989), or late adulthood for fossilization (MacWhinney, 2005). However, recent research (Hakuta, Bialystok, & Wiley, 2003; Wiley, Bialystok, & Hakuta, 2005) has cast doubt on many of these claims (MacWhinney, in press).

Despite these recent challenges, educators, academics, and the general public continue to believe in the reality of some Critical Period. What makes the notion of a Critical Period so compelling is that fact that adult second language learners often report problems acquiring a native accent in L2 and in using their L2 fluently. The approach to this issue that we have taken is to elaborate an extended version of the Competition Model (MacWhinney, in press) that accounts for age-related effects in second language learning through the mechanisms of entrenchment, transfer, and incomplete resonance. This new Unified Model makes strong predictions about the ways in which age-related effects can be overcome through effective teaching. In particular, the model holds that the problems that adults have in second language learning arise from the entrenched nature of the first language (L1), inadequate exposure to L2, and inappropriate teaching of L2. To correct these problems, teaching of adult learners needs to utilize these three methods: 1. Graduated interval recall, 2. Resonant cotraining, and 3. Cue focusing. The claim is that L2 instruction that incorporates these three methods will lead to marked improvements in fluency and robustness of learning.


graduated interval recall

This approach to the learning of items (words, sounds, constructions) in a second language was first elaborated by Pimsleur (1967), although components of the idea can be found as far back as Ebbinghaus (1885). Recently, Pavlik et al. (in press) have formalized the parameters controlling this procedure mathematically in the context of the ACT-R model of cognition. The core idea here is most easily illustrated in the context of the learning of a list of new L2 vocabulary items (Nation, 2001). Immediately after a word is presented, learners are almost always able to recall it. However, if we let a minute pass by, the memory trace drops below threshold and retrieval success drops with it. What Pimsleur discovered was that, if we retest the item before the memory trace decays too much, recall will be successful. Once an item has been recalled successfully once, repeated recalls trials can be spaced further and further apart. The neuronal basis of this process has now been elaborated in terms of synaptic reentry reinforcement model of hippocampal functioning (Wittenberg, Sullivan, & Tsien, 2002). Pavlik (in press) has shown that optimization of the intervals required for recall can lead to a two-fold improvement in vocabulary learning. This experimental work is now being extended to the in vivo study of online learning of Chinese vocabulary (Pavlik et al., in press) and pinyin dictation (Zhang, MacWhinney, & Wu, in preparation) in the PSLC online and offline courses. It has also been applied to the learning of Spanish vocabulary through a simple online tutor. The method of graduated interval recall is also being applied to the learning of French gender (Presson, Pavlik, MacWhinney, & Jones, in preparation). Each of these three efforts (Chinese vocabulary, Chinese pinyin, French gender) relies on the same code base for optimization developed by Pavlik.

resonant cotraining

The second mechanism for adult second language learning highlighted in the Unified Competition Model is the mechanism of resonant co-training. The basic effect of resonance can be most easily understood by contrasting the learning of French and Chinese. In French, learners have immediate access to a method for encoding the sounds of the language through Roman characters, including a few special French diacritics. Because adults rely so heavily on phonemic recoding during reading (Booth, Perfetti, & MacWhinney, 1999), they can easily form a resonant loop between a new auditory form and its meaning and orthography. This cortical loop serves as a scaffold for the process of hippocampal consolidation discussed above (Wittenberg, Sullivan, & Tsien, 2002). In Chinese, learners cannot form this loop, because they do not yet know most of the Hanzi characters required for writing and reading Chinese. It is this lack of orthographic resonance that makes Asian character-based languages like Japanese and Chinese so challenging for learners with a background in Roman characters. To improve resonance during learning, our systems for vocabulary learning interweave trials using pinyin, meaning, auditory form, and characters. Initial results show a significant advantage for training that incorporates this type of resonance. The PSLC Chinese project organized by Liu, Perfetti, and colleagues further elaborates in the role of resonance and co-training in learning tones. Rather than replicating those studies here, we hope to build on their results as a part of an integrated approach to the design of instruction in Spanish, Chinese, and French.

explicit cue focusing

The third mechanism for adult second language learning highlighted in the Unified Competition Model is the mechanism of cue focusing. Cue focusing is currently at the center of work by Zhang and Wu on Chinese pinyin dictation and Presson and Jones on French gender learning. The Chinese tutor allows students direct access to minimal pairs that characterize the correct target form (with tones and letters) and the form they have entered. In this case, cue focusing is explicit on a perceptual level. The French tutor presents cues in a simple, declarative form (i.e. -ance indicates feminine). In terms of the debate regarding implicit and explicit learning, both forms of feedback represent explicit teaching. However, as MacWhinney (1997) argued, explicit teaching is only successful when the cues are extremely simple. Both of these tutors rely on this core principle.


In the theory of self-organizing maps (Kohonen, 1970), entrenchment arises in cortical areas as a result of the formation of connections between mutually active items. In PDP theory, these are members of "gangs" participating in "gang effects" or correlated activation (Rogers & McClelland, 2006). Li, Zhao, & MacWhinney (2007) show how entrenchment increases across epochs of training of a SOM during growth in the size of the lexicon. Once forms are entrenched, it is difficult to organize them in alternative ways. However, there are various computational frameworks that can account for the overlay of a secondary organizations on an entrenched L1 structure.


Whenever a learner can map a new L2 structure onto a previous L1 structure, analogical transfer is possible. This transfer can be either positive or negative. Positive transfer arises when L1 matches L2. Negative transfer arises when L1 only partially matches L2.

social disincentives

Infants and toddlers receive a wide range of incentives for language learning, including gaze reinforcement, echoing imitation, games, smiles, and positive verbal input. Across the lifespan, these features mostly decrease until adolescent and adult language learners are eventually faced with many disincentives for language learning, including failure to respond in L2, negative feedback, and avoidance of interactions.

Research questions

Adult second language learning, unlike first language acquisition, must deal with learning barriers produced by L1 (first language) entrenchment, transfer, and social disincentives. In order to overcome these barriers, adult learners can rely on specialized reconfigurations of learning methods used by children learning their first language. These supports include: (1) graduated interval recall, (2) resonant co-training, and (3) explicit cue focusing. Presence of only one or two of these supports will lead to good learning, but the best and most robust learning occurs when all three are operative. This means that the overall hypothesis cannot be evaluated by a single definitive experiment. Instead, a series of experiments must be run to evaluate various configurations of the components. Also, it is possible that the effects of these methods may vary across linguistic levels (phonology, orthography, reading, lexicon, syntax, pragmatics, fluency). However, evidence for the effects of any combination of these supports in achieving any level of robustness on any given level would still provide important clues regarding ways to enhance the overall robustness of second language learning. This information could also be useful in understanding robustness in other domains.

Study One


We hypothesize that learners will benefit from the use of explicit cue marking for the correct formation of verbs in the 6 person-number combinations of the present, imperfect, and preterite of the verbs taught in first year Spanish. The control group will receive equivalent training, but without explicit cue identification.

Independent Variables

Group A will receive explicit cues for the preterite and subjunctive and non-explicit cues for the present, imperative, and imperfective. Group B will receive the reverse.

Dependent Variables

The dependent variables are response accuracy and response latency.


This study will begin in a pilot form at Penn State in the Spring. The target student population includes 900 subjects.


Study Two


We hypothesize that learners will benefit from the introduction of a graduated recall deadline for the correct formation of verbs in the 6 person-number combinations of the present, imperfect, and preterite of the verbs taught in first year Spanish. The control group will receive equivalent training, but without explicit cue identification.

Independent Variables

Group A will receive graduated deadlines for the preterite and subjunctive and no deadlines for the present, imperative, and imperfective. Group B will receive the reverse.

Dependent Variables



Further Information

Connections to Other Studies

Annotated Bibliography


Future Plans