1. ERPs in language acquisition
Judith Rispens
25 September 2009

2. Contents of presentation
Introduction to EEG and ERP
Designing an experiment
ERP technique: what does it take to get an ERP component?
ERPs and language acquisition: what components are there to investigate?
Do’s and don’ts

3. EEG / ERP EEG: Electroencephalogram ERP: Event-related potential
Neuro-imaging method for investigating neurophysiological correlates of language processing
Different from fMRI and PET: high temporal resolution (but poor spatial resolution)
Non-invasive technique
Relatively cheap
Can be used with infants, older children and adults

4. Electroencephalogram
Record of electrical brain activity over time
EEG constituted by neurons that polarise or depolarise simultaneously
Activity measured at the scalp

6. Event-related potential: response to specific external stimulus
Difference in voltage small, so many trials (+/- 50 trials) needed
Signal is amplified
These are averaged and extracted from the background noise of the ongoing EEG
Event-related brain potential
An estimate of the time-locked neural activity evoked by an external stimulus

7. Event Related brain Potentials
Properties of ERPs
Time-locked to external stimulus
Negative or positive peak
High temporal resolution (in ms)
Measured at different positions at the scalp
Scalp distribution of ERP components, but inverse problem: do not know the origin of the measured activity at the scalp
fMRI research related to ERP components gives insight into localisation
Techniques for spatial modelling (e.g. LORETA)

8. Event-Related brain Potentials
Different linguistic stimuli elicit different ERP components
Named after the polarity (negative / positive), latency, appearance at the scalp, or its function
N400, P600, Early left anterior negativity (ELAN), mismatch negativity (MMN) etc.

10. Event-Related brain Potentials
Properties of ERPs:
Latency (in milliseconds)
Onset of ERP component
Peak latency
Amplitude (in microvolt)
Strength of component: indication of processing load
Distribution
At the scalp (but be aware of inverse problem)

11. Tools ± 50.000 € What does it take to get an ERP component?
ERP software, e.g. Neuroscan
Software to present stimuli, e.g. E-Prime
Cap with electrodes (e.g. Easycap)
Amplifiers
Computers
Subjects
Stimuli
Response box
Headphones / speakers
± €

12. Tools How does it work? Prepare scalp Takes around 15-30 minutes
‘scrub’ it with alcohol
Put on cap
Apply gel for conduction
Apply electrodes
Electrodes to measure eye movements / blinks
Try to keep impedance less than 5 kohm
Takes around minutes

14. Experimental design Presentation of stimuli:
Code the conditions, items and critical sections within your items so that you can track in the ongoing EEG the different responses to the different stimuli
Carry out the experiment and record the ongoing EEG
After the experiment you remove responses containing artefacts like eye-movements, blinks and head movements as this gives noise in your data
Average stimuli of a similar condition
May apply a filter to ‘clean up’ your data

15. Designing an experiment
Think about:
Auditory or visual presentation of stimuli
Do you need a response to your stimuli (e.g. judgements)?
If yes, make sure that the behavioural response does not interfere with the EEG recording
Position of condition label within a word/ sentence
* Hij heeft een gelezen
Filler stimuli
Around items per condition
Appropriate items for children (e.g. check age of acquisition)
Position of target word in context: wrap up’ effect at sentence end may interfere with ERP component, better to have target mid-sentence

16. Designing an experiment
Subject selection
Minimum of 20 subjects
Are demands on subjects feasible?
Children are less able to sit still and trials with muscular movement need to be removed.
Control for age, sex and handedness (in infants parent’s handedness)
Children: standardised language tests to check language level in case of suspicion (> 2 years old)
Children: screen family background for genetic risk of developing language impairment such as developmental dyslexia and SLI
Spend time on subject recruitment: more difficult than behavioural experiments

17. ERPs and language acquisition
Language related ERP components
Mismatch negativity
ELAN
N400
P600
Keep in mind:
Properties of ERP components may differ between adults and children due to physiological differences
Makes it difficult to make comparisons between adult and child studies

18. ERPs and language acquisition
Mismatch negativity (auditory perception)
Index of the detection of a change in an acoustic signal. Can be used with non-speech and speech sounds
Presented in ‘oddball’ paradigm: frequent and infrequent stimuli presented in a 9 : 1 ratio
The deviant stimulus elicits the mismatch with the neuronal memory trace
Between ms after stimulus onset negativity at frontal sites
Elicited without attention at stimulus
No behavioural response is needed thus suitable for infant studies

19. ERPs and language acquisition
MMN
F2 electrode

20. ERPs and language acquisition
Why would you study MMN?
Investigate phoneme boundaries, auditory perception of sound qualities such as duration, pitch etc.
Manipulate difference between sounds
Present 90% of cases stimulus 1 and 10% stimulus 2
If infant discriminates between the two sounds, MMN is visible
For infant and child studies: the MMN is different from that of adults. MMN preceded by an early positivity

21. ERPs and language acquisition
Mismatch negativity
Example of bak-dak experiment
NWO project: investigating early precursors of developmental dyslexia
Goal: investigate phoneme boundaries in infants of 8 weeks old at genetic risk of dyslexia.
Hypothesis: auditory processing is compromised in developmental dyslexia

22. bak: - b-a dak: \ d \ a
Van Beinum et al. (2005).

23. 1 2 3 4 5 6 7 8 9 10 From BAK to DAK bak 1 bak 2 bak 3 bak 4 bak 5

24. ERPs and language acquisition
Child listens for +/- 45 minutes to stimuli while sleeping

25. 8 weeks old
Control babies
At-risk babies
Van Leeuwen et al. 2006

26. ERPs and language acquisition
Early Left Anterior Negativity (syntax)
Around 150 ms after violation onset
Violation of syntactic phrase structure elicits ELAN
*Het meisje heeft in de geslapen
*The girl has in the slept
Automatic process (independent of task demands and attention)
ELAN reflects highly automatic phrase structure building process

27. ERPs and language acquisition
ELAN
Example of experiment (Hahne et al., 2004)
Goal investigate developmental pattern of ELAN
Subjects: 6, 7, 8, 10 and 13 year old children
Materials: a phrase structure error
* the goose was in the fed (incomplete PP)
The goose was fed

28. Results 6 year olds: no ELAN 7-10 year olds: sustained ELAN
13 year olds: ELAN comparable to adults

29. ERPs and language acquisition
Conclusion:
Conclusions
Automatic adult-like structure building established quite late (13 years) in development and absent in early development (until 6 years)
Question: sustained negativity a precursor of ELAN or does it reflect different processing mechanism?

30. ERPs and language acquisition
N400 (lexical semantics)
Negative wave after 400 ms of stimulus presentation most prominent at centro-parietal sites
An index of semantic processing (e.g., Kutas & Hillyard, 1980)
Semantically incongruous stimuli elicit greater N400 amplitudes than semantically congruous stimuli
This reflects a greater effort at semantic integration for unexpected than expected stimuli
The pizza was too hot to read

31. ERPs and language acquisition
Example of experiment
Friedrich & Friederici (2005)
Goal: To investigate whether adult-like mechanisms of semantic integration, as indexed by N400, are present in 19-month-olds
Task: picture-word-matching
Congruous condition: picture-word match
Incongruous condition: picture-word mismatch
Subjects: month-old German children

32. ERPs and language acquisition
“ball” (congruous)
“duck” (incongruous)
“A…”
Picture
onset
Indef.
article
Target
word
0 ms
1000 ms
2000 ms
3000 ms
4000 ms

33. ERPs and language acquisition
Results
A broadly distributed long-lasting semantic incongruity effect
ms: incongruous words more negative than congruous words
At centro-parietal & frontal sites
Possibly more contribution from the left hemisphere
The spatio-temporal distribution of the semantic incongruity in 19-month-olds differs from that in adults.

34. ERPs and Language acquisition
Associated with syntactic reanalysis or revision
Is triggered by (morpho-)syntactic violations and garden-path sentences
Is affected by task demands, attention and frequency of violation: keep in mind when designing your experiment!!
Centro-parietal distribution

35. P600

36. P600 Example of experiment Oberecker, Friedrich & Friederici (2005)
Are the properties of the P600 component of 32 month old children comparable to those of adults?
Latency, amplitude, distribution
Experimental conditions:
The lion roars (control)
The lion in the roars (violation)

37. P600 Experiment failed…. Too many sentences (250)
Too many blinks/movements
Delayed response interfered with interpreting the behavioural data

38. P600 Experiment failed…. Too many sentences (250)
Too many blinks/movements
Delayed response interfered with interpreting the behavioural data

39. P600 Results: Conclusions:
Different from adults as young children show very early positivity
P600 present but with delayed latency
Conclusions:
2.8 year old children comprehend simple sentences with a neural system similar to adults. The delayed latencies is taken to reflect a developing system

40. Some points of attention
ERP infant research troubled by different sleep stages, sucking movements etc.
ERP infant research troubled by movements, blinks etc.
ERP children research challenged by the demands of sitting still
Reliability of judgement data
Infant skull is still growing and changing, makes it difficult to interpret data longitudinally
ERP data in general unreliable at the individual level

41. Practical tips
Ask parents to measure their children’s head before coming to the lab, so you can prepare the right size cap
Ask children not to wear fleece sweaters
Ask participants to have washed their hair before coming (no gel etc.)
Ask participants not to wear expensive clothes as conduction gel may stain
Let children watch cartoons during cap preparation
Make a picture with a digital camera and give a print-out before leaving

42. Goodluck with your experiments!!!!!!!