Writing Assignment 2

Writing Assignment 2

Keep your writing to two or three pages.Illustrations (e.g. diagrams, figures, and tables) are acceptable only if they enhance the interpretability of the paper. They do not count toward the page limit.Your written assignment has to be submitted in electronic form (preferably in Word; PDF is acceptable) to the Electronic Dropbox.Handwritten assignments will not be accepted.

If you would like to collaborate with other students (i.e., turn in the same text for a group of students), that is acceptable. Please keep your collaboration to two or three students (including yourself). If you do collaborate, write the assignment together and put all your names plus student ID’s on the assignment. Also, each student should separately submit the assignment to the Dropbox.

Interactive Activation Model (IAC)

For this assignment, you will be using a java applet that simulates the IAC model:

After you start the applet, click on “Get Lexicon” to load in lexicon consisting of 1323 English 4-letter words. Note that the model only knows about these 4 letter words.

Figure 1. A illustration of the IAC model. Note that only a small sample of word, letter and feature nodes are shown – for each letter of a four letter word, there are separate feature and letter nodes and in this illustration, only a few letter nodes for the third position are shown. Lines with arrows and circles indicate excitatory and inhibitory connections respectively.

In the input box, you can write a four letter word. The letter features of this word will be presented as input to the model. For example, “HIRE” will activate the features for the letter H in the first position, the letter I in the second position, etc. It is also possible to leave letter features blank by providing an underscore. For example, “_IRE” will not activate any letter features for the first letter. Finally, you can use an asterisk to represent an ambiguous letter that resembles both an R and a K. For example, “WOR*” will activate both the letter features of R as well as K in the fourth letter position.

By clicking “RUN”, you can simulate the model by passing activation to the feature nodes for a number of iterations specified in the #cycles box. The applet creates a graph showing the activations of nodes as a function of time (i.e., cycles). You can create graphs for both word nodes and letter nodes (use the layer selector to switch between word and letter nodes). When you select letter nodes, you can switch between letters at the first, second, third, and fourth positions.

Please address the following points in your writeup:

A) Present the string “WORD” to the model. Which word node becomes most active? Which word node is associated with an initial rise followed by a decrease in activation? Can you explain why this happens?

B) Present the string “WOR_” to the model. Which word nodes become most active? Which letter nodes in the fourth position become most active? Can you explain why letter units become active in the fourth position even though no activation is passed bottom-up for the feature units in the fourth position?

C) Present the string “WOR*” to the model. What happens now to the word nodes and letter nodes in the fourth position?

D) Present the string “MAV_” to the model. There is no fourth letter that makes this string a valid English word. What happens to the letter nodes in the fourth position? Can you explain this result?

------

Note: for additional background information, read the article (but this is not required)

McClelland, J. L. & Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: Part 1. An account of basic findings. Psychological Review, 88, 375-407.

This article can be found here: