Electrodermal activity (EDA)

What does this channel measure?

EDA measures the subject's skin conductance (sweat gland activity, sometimes called galvanic skin response). ANSLAB identifies significant changes in skin conductance level defined as drops or rises greater than .02 microSiemens above a zero slope baseline within a defined time interval. These are called SCRs (skin conductance responses) and, depending on the experimental context, can be specific or non-specific (without known stimulus).

SCRs can be mapped to experimental events to measure subject reactivity to various stimuli. Non-specific SCRs can be counted per unit of time (e.g. 1 min) and are called non-specific fluctuations (NSFs).

If you are interested in evoked SCRs, you first scan the entire file for artifacts and correct them if these are found in a sensitive position. Then, stimulus-related segments are extracted and amplitude changes can either be scored trial by trial or averaged for every sample point across experimental conditions.

For non-specific fluctuations, the procedure is different: after editing artifacts, fluctuations are recognized automatically and displayed for you to check for correctness. Both types of processing are described separately in the following sections.

[Top]

Different processing types

1. Evoked skin conductance responses (SCR)

If you are analyzing an event-related design, the objective of scl-editing is not to extract certain calculated traces but to identify and edit artifactual data segments. The 'clean' signal can then be used for automatic extraction of event related epochs (trials) as described on the Analyzing event-related designs-page.

Therefore analyzing involves only 'resuming' the analysis once after artifact editing in order to save the clean signal.

You can load optional channels (e.g. accelerometer or marker) to give you more information about data segments you wish to edit. Artifacts are removed by using the buttons in the 'editing section' of the command window, especially the 'connect'-function to remove and interpolate.

The most common artifacts are technical errors in hookup or data collection. These artifacts will appear as large spikes or sudden niveau changes in the EDA graph.

Another irregularity that must be noted but cannot be edited is the non-responder or person with little electrodermal lability. A small percentage of the general population do not respond electrodermally. A non-responder can be recognized by a restricted range on the vertical axis and the lack of identified changes >.02 uS. One needs to decide a priori whether to exclude the data from non-reponders (typically data is included).

2. Non-specific fluctuations (NSF)

Extracting non-specific fluctuation parameters of the skin conductance level also involves a first step for artifact editing. The accelerometer signal can be loaded also to help identifying movement artifacts (by specifying an optional channel to load).

When editing is finished, hitting the resume button will start the response identifaction algorithm. Before running the fluctuation detection, ANSLAB filters the edited signal with the given cut-off frequency in the eda-options, to avoid false positives due to high frequent noise. This is not done directly after loading the data, as artefacts would then be smoothed out and would be nearly impossible to identify.

Detected SCR onsets are then displayed in the main axis and qualifying onsets (baseline points) are labelled with small numbered red dots. These dots are only plotted when zoomed in sufficiently, to avoid long drawing times when browsing large data files.

Two additional axes display bivariate distributions of SCR pattern characteristics. Typically outliers are due to technical or movement artifact. These need to be inspected and excluded if necessary. You can automatically jump to a specific response by clicking on the corresponding number on the two bivariate plots on the right.

Besides the skin conductance level, the following variables are extracted from the electrodermal activity channel: These can be shown by switching from 'raw' to 'event'-mode:

You can set artifactual intervals to NaN (missing data) by using the 'exclusion box'-button or define a global artifact using the 'define artifact'-button.

Once you are done, push the 'Resume'-button and select 'Save reduced data' if you want to save your changes.

The following figure shows the SCR-rise time and the SCR-half recovery time of the skin conductance level. Red dots with numbers (16, 17 in the upper case) represent qualifying onsets (baseline points)


3. Single trial scoring

This method allows to score single skin conductances responses, e.g. following a stimulus used for fear conditioning.

Prior to the scoring you need to create a timing file for each raw file which contains just the segments inside which you want to score skin conductance responses. If you edited the underlying EDA data already before you may also choose to use the already edited data (see the options to enable/disable this option).

Once you load a file, ANSLAB will go through all the segments found in the according timing file. For each segment ANSLAB detects the minimum and maximum of the skin conductance level. The detection of the minima and maxima is restricted to the response window as configured in the options. So the timing files are only used to obtain the stimulus onset for each trial.

Scoring is then done using the dynamic section of ANSLAB. If the response looks fine, press 'accept. Use the buttons 'invalid' and 'zero' if the EDA response is invalid or if there is no response at all, respectively. You may also adjust the response window (for the detection of the minimum and maximum) and the y-range of the graph showing the skin conductance level using the controls in the dynamic section.

After scoring all segments ANSLAB will store a .mat file containing the scoring data, a text file for the scored subject which also contains information about each segment scored, and a text file containing scoring information for all subjects. All these files are stored within the subfolder 'scoring' inside the EDA folder of your study folder.

[Top]

What kind of artifacts are common in this channel?

ANSLAB almost always identifies significant SCRs correctly, so this channel does not require heavy editing.

ANSLAB will occasionally fail to identify significant changes in SCL or will misidentify non-significant changes. Scan the graph and note on the vertical axis the height of all rises and dips. If you suspect that ANSLAB has made a mistake, follow the instructions below to add or remove dots.

[Top]

How are artifacts removed?

You mainly need to look out for movement artifacts that become apparent in the bivariate plots or for instances when ANSLAB has missed or misdetected an SCR. Before every wave (>0.02 uS) in the signal there should be a red dot marking that the program has detected this event. If there is no such dot, it might be that the rise was too slow, that the amplitude was less than 0.02 uS, or that the signal was very noisy at this point.

If you don not agree, you need to push the 'insert'-button on the command window. Then you can click on a position where you wish to insert an SCR. You can delete events with the 'delete'-button and clicking near the event you wish to delete. The bivariate plots are updated to include your changes, allowing you to check your editing results. You can also use the 'delete box'-button to delete multiple events at a time.

[Top]

What qualities must be preserved in editing?

The most important editing goal is to remove obvious technical artifacts that distort a mean. The second goal is to identify electrodermally stable individuals so that they can be considered separately or in some cases excluded from data analysis.

[Top]