The EEG structure

Why do I need this step?

You will use the same dot syntax you used with the demo variable in the MATLAB basics lesson: EEG.srate means the srate field inside EEG.

How do I do it in EEGLAB?

1. Open the fields in the Workspace

1. Open the dataset as in the previous lesson.
2. In the MATLAB Workspace panel, double-click EEG.
3. The fields of EEG appear in a table. Double-click any field to open it, such as chanlocs or event.

2. Read the events from EEGLAB

1. Open Edit › then Event values.
2. A window shows one event at a time: type, position and latency. (The urevent field is not shown in this window.)
3. Notice that the latency field in this window is titled latency (sec): the window shows the latency in seconds, while EEG.event stores it in samples. The first event appears here at about 1.0001 seconds.
4. Close the window with Cancel so you do not change anything.

3. Read the dataset information

Open Edit › then Dataset info. Values such as the dataset name and sampling rate appear. Close it with Cancel.

How do I write it as code?

Type these lines in the Command Window, one at a time, and compare the output with the comment:

% Basic information
EEG.setname            % 'Continuous EEG Data'
EEG.nbchan             % 32 channels
EEG.srate              % 128 samples per second
EEG.pnts               % 30504 samples per channel
(EEG.pnts - 1) / EEG.srate   % 238.3047 seconds (same as EEG.xmax)

% The signal: channels x samples
size(EEG.data)         % 32 30504
class(EEG.data)        % 'single'
EEG.data(1, 1:5)       % first 5 samples of channel 1 (microvolts)

% Channel names
{EEG.chanlocs(1:3).labels}   % 'FPz' 'EOG1' 'F3'

% First event: type and latency
EEG.event(1).type      % 'square'
EEG.event(1).latency   % 129.0087 (in samples)
(EEG.event(1).latency - 1) / EEG.srate   % 1.0001 (in seconds)

% Empty until ICA is computed
isempty(EEG.icaweights)   % 1 (true)

• Why subtract 1? In MATLAB the first sample is number 1 and sits at time zero. So time in seconds = (sample number − 1) ÷ sampling rate.
• EEG.data(1, 1:5): the first number selects the channel (row), and the second selects the samples (columns).

Which value should I choose?

There is no value to choose here. These are the main fields you will come back to in every lesson:

How do I check it worked?

1. size(EEG.data) gives 32 30504.
2. (EEG.pnts - 1) / EEG.srate gives 238.3047, the same as EEG.xmax and as Epoch end (sec) in the main window (shown rounded as 238.305).
3. EEG.event(1).type gives 'square', and the converted latency gives about 1.0001 seconds, matching what you saw in Edit › then Event values.
4. numel(EEG.event) gives 154, matching the Events line in the main window.

What mistakes do people make?

• Swapping the dimensions. In EEGLAB channels are rows and samples are columns. EEG.data(:, 1) is all channels at the first sample, not the whole first channel. The whole first channel is EEG.data(1, :).
• Reading latency as seconds. In EEG.event the latency is in samples. Convert it with the formula above, and do not forget to subtract 1.
• Forgetting that EEG.times is in milliseconds. While EEG.xmin and EEG.xmax are in seconds.
• Editing EEG by hand and not telling EEGLAB. If you change a field yourself, run EEG = eeg_checkset(EEG);, then store it with [ALLEEG, EEG, CURRENTSET] = eeg_store(ALLEEG, EEG, CURRENTSET); and eeglab redraw. Otherwise the main window may show outdated values.
• Mixing up channel and electrode. EEG.chanlocs(5) is the fifth channel in the data, and its position is the position of the electrode that recorded it. The channel order in a file does not follow any order on the head.

Going further

• EEGLAB Tutorial: https://eeglab.org/tutorials/ConceptsGuide/Data_Structures.html (EEGLAB Data Structures: EEG.event, Event latencies)
• EEGLAB Tutorial: https://eeglab.org/tutorials/02_Quickstart/quickstart.html (Exploring event values)