Theories of Language Learning

The Cognitive Perspective

Interface of linguistics and psychology, study of language in the mind and in the brain. Growing influence of brain research in the last decades that leads us to experiment new approaches to the study of language. Checking brain activity during language processing.

- Cognitive linguistics, related to usage-based theories of LA

- Psycholinguistics, experimental techniques to test language processing (Information Processing theory)

- Neurolinguistics, (study by Paradis with aphasics)

- Cognitive neuroscience

New interests connected with the cognitive resources needed for SLA. The role of:

- memory

- attention

- automaticity and automatization

- sentence processing

- motivation

- learnability

Why is this cooperation so important? Language studies are usually focused on output, the product of the process of language, but we never check how this output is generated and the process of perception. The major goal of linguists has been to describe grammar, to provide descriptive rules and

Patterns of the language system, without any concern for psycholinguistic validity;

Psychologists have focused on the mental processes and structures, with little concern for the linguistic patterns of the language code.

It is necessary that these two work together to produce valid results.

Not only what the learner knows but also what he does.

The anatomy of the brain

Besides Broca's and Wernicke's areas there are several other regions of the brain that contribute to language processing (even parts of the right hemisphere).

Arguments supporting the existence of a functional anatomy of the brain:

• Localization argument: there are specific areas for every pattern of language
• BUT there isn't a universally accepted, detailed language specific topography of the brain (Area for syntax? Area for phonetics? Etc.) because:
• - cognitive neuroscience is relatively new and still in development
• - the failure to identify specific areas may be due to the fact that some language

Functions aren't correlated to specific cerebral regions

Arguments against the existence of a functional anatomy of the brain:

• Anti-localization argument (Baars, 2007): certain parts of the neuron system are so heavily interconnected and the constituents interact with each other so strongly that we cannot isolate individual functional modules.

Research methods for investigating language and the brain:

Psycholinguistic methods:

• RTs studies
• Priming
• Self paced reading
• Eye tracking
• Computational modeling

Neuroimaging:

• Electrophysiological methods (EEG, ERP)
• Haemodynamic methods (fMRI, PET)

Tasks:

Offline tasks → used to measure learner knowledge, often involve paper-and-pencil type tests

• grammaticality judgment tests (learners indicate whether they believe a sentence is possible or not in a given language)
• truth value judgment tests (learners determine which of two sentences logically follows or fits with something they just read)
• cloze tests (learners

1. Online tasks → moment-by-moment measures of what learners are doing with language. Tend to measure underlying processes
2. Moving window technique (learners read bits and pieces of a sentence on a computer, push a button → screen, advancing through the sentence as they self-paced reading) (the movement of people’s pupils is tracked while they read a sentence on a computer screen)
3. Eye tracking
4. Measuring of RTs (longer reading times = longer processing times) – Neuroimaging. Isn’t very substantial.

Contribution to the study of language:

Is a very recent development, used in language-related research from the 1990s.

Methods can be divided into 3 groups:

1. CT scanning (X-ray technologies)
2. Electrophysiological methods
3. Haemodynamic methods

It can be used for two different purposes:

• Structural imaging that produces static pictures of the structure of the brain (not very useful)

Functional neuroimaging involves measuring some aspects of brain function in order to understand the mental operations associated with it (images of the brain in action).

Electrophysiological methods, such as Event Related Potentials (ERPs), reflect the online activity of groups of neurons and measure electrical neural activity. They allow for the recording of neurological activity associated with a specific cognitive task (stimuli).

We need to compare the different conditions to the baseline condition. However, ERPs cannot be recorded during natural speech production because the neural activity involved in coordinating articulatory muscles would interfere with the results.

The output of ERPs is a waveform that records negative (N) and positive (P) peaks relative to the baseline. The most stable observed effects are:

• N400 semantic violation effect
• P600 syntactic violation effect

The N400 component is the most studied language-related component.

refers to Event-Related Potentials, which are electrical brain responses that are measured using electroencephalography (EEG). ERPs provide information about the timing and processing of cognitive events in the brain. The N400 component is a negative voltage peak that occurs around 400ms after a stimulus. It is sensitive to various lexical properties such as frequency, familiarity, and integration within the context. The position of the stimulus must always be the same. The largest source of the N400 component is the left temporal lobe, with a lesser contribution from the right temporal lobe. An example of the N400 component is observed when a sentence contains a word that is semantically incongruent, resulting in a graded increase of the N400 component. The P600 component is a positive peak that occurs 600ms after the onset of a critical stimulus. It is associated with syntactic violations or syntactic complexities. The P600 component is interpreted as reflecting processes of reanalysis and syntactic repairs. These methodologies, ERPs, have good temporal resolution (they tell us when an event occurs), but they lack spatial resolution (they don't tell us where in the brain the activity is happening) and do not provide information about the nature or reason behind the brain's reaction.SLA research

Are there qualitative and quantitative differences in the manner in which L1 and L2 speakers process the language?

• Qualitative differences: responses of a different type or even absence of a specific component
• Quantitative differences: delay in time or in the size of a signal change

These methodologies have been used in 3 domains:

1. L2 perception and sound processing (more on native and nonnative contrasts, not SLA)
2. Lexical processing
3. Sentence processing

1. ERP and sound perception

When combined with a behavioural task, neurophysiological discrimination can transpire before or without behavioural discrimination.

Is there a correlation between the ability to discriminate a contrast behaviourally and the electrophysiological effect? → native vs nonnative contrast: an electrophysiological effect was generated by both types of contrasts, but only the native contrast was discriminated behaviourally.

This could be due to the fact that they are discriminated at a different

level of sound processing. The difference between two sounds in a native contrast is perceived as being linguistically meaningful, BUT the difference between two sounds in a nonnative contrast is sometimes perceived as being purely acoustic.

Study on Finnish and Hungarian learners of L2 Finnish on two vowels contrasts. These have been discriminated by all groups, not only behaviourally, but there was also an electrophysiological response.

The results suggested that, given enough exposure to the language, it is possible for L2 learners to develop new memory traces for L2 sounds even when the acquisition starts later in life!

When we are exposed to a contrast NOT of our L1, but proper of our L2, we have an electrophysiological response. Are we CONSCIOUS of the effect? (Behavioural response?)

Behavioural means that I have to decide something. Electrophysiological means that I perceive something different. Relationship between these results and behavioural results?

2. ERP and lexical processing L2

1. Processing of anomalous lexical items results in native-like N400 effects with an occasional quantitative difference.
2. No matter how late the L2 was learned, an N400 effect is generally seen in L2 lexical semantic processing.
3. Thus, there seem to be no major qualitative differences in how lexical semantics are processed in L1 and L2.

→ Haemodynamic methods.

They map the changes in the blood supply of neurons. Increased neural activity requires a local increase in the cerebral blood flow to provide the necessary energy.

• It is very accurate in terms of spatial resolution.
• The temporal resolution is nowhere near as good as the millisecond-level precision of electrophysiological methods.

→ fMRI results suggest very similar processing regions for L1 and L2, with extent of activations depending on proficiency and AoA. PROFICIENCY seems to affect processing more than AoA.

Psycholinguistic methods.

Have been used for FLA for many years.

They are behavioural tasks (not the brain activity, but the cognitive processes involved in language learning).

actual reaction, the actual behaviour of the subject). Subjects are presented with a task, a linguistic stimuli, and they have to perform an action. Some of these tasks are: - lexical decision - picture naming Measures of performance: - reaction times, on the basis of which researchers can make inferences about the nature of the processing involved during the task; - error rate, the proportion of correct responses. What is the NATURE of the observed effect? Interpretation. There are 3 main methodologies: 1. Self-paced reading (for syntax) - using a pc, subjects are asked to read a sentence in a word-by-word or phrase-by-phrase fashion by pressing a button to request the next word/phrase to appear on the screen; - it is necessary to figure out ways of controlling the subject during the experiment (to check if he is really paying attention to the sentence), so each sentence is followed by a yes/no or true/false comprehension question; - there are always two sentences, one correct and one

incorrect (baseline conditions);

• importance of memory;
• slower RTs when we expect a certain word, when there are strong correlations between words;
• longer RTs in presence of processing difficulties;
• potential problems: usually slower than normal reading pace are recorded, the need for the reader to remember the previous parts without the option of going back to reread words may result in short-term memory interference. (eye doesn't have these problems)

2. Eye tracking

• records the movement of the eyes while reading very accurately:
• it is also useful to examine general mental operation through examining visual attention;
• we do not read continuously, but we skip through parts of the sentence (saccades).