In contrast to the interactive activation model, in which the letter layer is connected directly to the word layer, in the spatial coding model the letter layer sends inputs to the word layer via a spatial code layer. Spatial coding is a form of orthographic input coding in which a two-dimensional code is used to encode information about both letter position and the degree of evidence for that little. We can get a better idea of what the spatial code looks like by plotting the state of that layer during a trial. To do that, right click on the spatial code layer and choose plot state.

Now let's run a trial. There are a few different types of trials that come preloaded with the spatial coding model; for now, let's choose an unprimed lexical decision trial. To do that, select ldt from the box labelled Trial in the control bar. Type form in the box labelled Stimulus and then click the Run button. Then go to the Results tab to look at the plot. It should look like this:

Each panel in this plot represents a different point in time during the trial, starting at the top left and moving left to right, top to down over time. Each of the Gaussian-shaped functions represents the signal of the different letter node. The position on the location axis reflects letter position information, while the amplitude reflects the strength of the signal, which increases over time. The key thing to note for present purposes is that there is uncertainty associated with letter position. This means that the letter position signal spills over somewhat to adjacent letter positions, which is central to the model's account of transposed letter similarity, for example.

Next: Plotting match values