Now let’s see how event codes are assigned to bins with BINLISTER, using the very simple analysis shown in Figure 6.2 in which we have one bin for the Rare category and another for the Frequent category. We’ll exclude trials in which the buttonpress response was incorrect.
As you’ll recall from Chapter 2, a bin descriptor file is used to tell BINLISTER how event codes should be assigned to bins. In the Chapter_6 folder, you’ll find a bin descriptor file named BDF_P3.txt, which we’ll use for this exercise. Make sure that the Chapter_6 folder is the current folder in Matlab, and double-click the BDF_P3.txt file from the Current Folder pane in Matlab to open it. Here’s what you should see:
bin 1
Rare, Correct
.{11;22;33;44;55}{t<200-1000>201}
bin 2
Frequent, Correct
.{12;13;14;15;21;23;24;25;31;32;34;35;41;42;43;45;51;52;53;54}{t<200-1000>201}
Each bin is described by a set of three lines. The first is the bin number (which must be in consecutive order, beginning with 1). The second line is the label for the bin (which can be anything you like). The third line is the actual bin descriptor. A bin descriptor indicates the sequence of event codes that define the bin. Each set of curly brackets ("{}") defines an event list that contains one or more event codes. For each bin descriptor, one event list must be preceded by a period symbol. This event list defines the time-locking event for the epoch (i.e., time zero). In the example shown above, event codes 11, 22, 33, 44, and 55 will serve as the time-locking event for Bin 1. This bin therefore includes trials with an A stimulus when A is the target, B when B is the target, C when C is the target, D when D is the target, and E when E is the target. We could have instead created a separate bin for each of these five target letters and then combined the five bins after averaging using ERP Bin Operations. However, it was simpler to combine them at the BINLISTER stage. The event list for Bin 2 contains all the event codes for the nontarget letters. You can verify this by comparing the event descriptors with the list of event codes in Table 6.1.
The time-locking event list may be preceded or followed by other event lists, indicating that those events must be present for an epoch of EEG to be assigned to a given bin. For example, imagine that Bin 1 was defined as:
{202}.{11;22;33;44;55}{201}
201 is the event code for a correct response and 202 is the event code for an incorrect response, so this bin descriptor would find targets (event codes 11, 22, 33, 44, and 55) that are immediately preceded by an incorrect response and immediately followed by a correct response.
In the actual bin descriptor for Bin 1, we don’t require any particular event code prior to the time-locking event, but we do require that the time-locking event (the stimulus) is followed by the event code for a correct response. However, we want to make sure that the response time (RT) wasn’t an outlier, indicating either a fast guess (an RT of <200 ms) or poor attention (an RT of >1000 ms). To do this, we use a time-conditioned event list in which the list of event codes is preceded by t<start–end> (e.g., t<200–1000>201 to indicate that event code 201 must be 200-1000 ms after the time-locking event).
Note that if we didn’t use a time-conditioned event list and instead used .{11;22;33;44;55}{201} as the event descriptor, the response event code (201) would need to directly follow the stimulus event code, with no other event codes between. However, by using a time-conditioned event list to specify that the 201 must be between 200 and 1000 ms after the stimulus event code, other event codes may occur between the stimulus and the response.
Also, if a time-conditioned event list appears prior to the time-locking event, time flows backward from the time-locking event. For example, if you specify {t<200-800>15}.{100}, BINLISTER will search for an event code of 100 preceded by an event code of 15 that occurred 200-800 ms prior to the 100. Additional details can be found in the BINLISTER documentation.
Should you exclude trials with incorrect behavioral responses?
In this experiment, we excluded trials with incorrect behavioral responses. In other studies, however, we don’t exclude the error trials. Here’s the general principle: If the errors in a given task are likely the result of lapses of attention, then exclude the error trials; if the errors in a given task mainly occur because the task is very difficult, then don’t exclude the error trials.
For example, errors in most oddball paradigms are typically a result of lapses of attention, so we exclude the error trials. Indeed, the ERPs are quite different on error and correct trials in the oddball paradigm (Falkenstein et al., 1990). By contrast, in tasks using the change detection paradigm to study visual working memory, most errors occur because of limits in storage capacity (Luck & Vogel, 2013). As a result, brain activity is similar on correct and error trials (Luria et al., 2016), so we don’t exclude the errors (which would reduce the number of trials per bin quite a lot, decreasing the signal-to-noise ratio).
