Learning & Teaching Foreign Languages

Chargement du moteur de recherche...

L2 past

Read and Reflect

Ellis and Schmidt conducted a second language connectionist experiment similar to Rumelhart and McClelland's landmark English past tense study.  These pioneering connectionists trained a virtual network on the past tense forms of regular and irregular, frequent and rare English verbs and after many repetitions, the network produced correct past tense forms for the training verbs, and was also able to generate correct forms for unfamiliar verbs, both regular and irregular, frequent and rare.

Ellis and Schmidt developed an artificial language to teach to both adult learners and a virtual network.  They also tested regular and irregular forms, encountered frequently and rarely, and measured how accurately and how quickly learners responded.  They found more errors and slower responses for less frequent items, regular or irregular, in early stages of learning, but only for infrequent irregulars at later stages.  They conclude that this
 
regularity by frequency interaction is a natural consequence of the power law of practice and is thus entirely consistent with associative learning processes: Regularity is frequency by another name.

(Ellis & Schmidt, 1998)


The virtual network produced the same learning patterns as their human subjects, providing strong support for connectionist accounts of language learning.
 
1. Why do you think the researchers used an artificial language instead of studying second language learners of a natural language?

2. How can we explain the finding that learners are slow on all infrequently heard verbs at the start of the experiment, but only on infrequent irregular forms later?

ANSWERS

1. Why do you think the researchers used an artificial language instead of studying second language learners of a natural language?

In natural acquisition circumstances, it is not possible to control the frequency of elements in the input: there will always be variation in the number of occasions different learners hear and use different verbs.

2. How can we explain the finding that learners are slow on all infrequently heard verbs at the start of the experiment, but only on infrequent irregular forms later?

When a verb is only rarely encountered, the connection between the base form and the past tense is weak, whether the past form is rule-based or not. But as more connections are made and strengthened as learning progresses, connections with regular past forms are validated more frequently and thus become stronger. This allows learners to retrieve the regular forms more quickly. Irregular forms which are met frequently also have strengthened connections. Only infrequent irregular forms have fewer connections and so remain difficult to retrieve.

Valid XHTML 1.0 Transitional