Language, Cognition & Computation

by Albert Major

How We Read – Cognitive and Computational Models

1. Reading is a very complex process

When we read, a lot of mental processes happen, one after another and sometimes at the same time (in parallel). These include:

• Eye movement: Moving the eyes from one word to another (this movement is called a "saccade").
• Visual feature extraction: Recognizing the basic visual shapes and letters of the word.
• Lexical access: Finding the meaning of the word from our mental dictionary (called the "lexicon").
• Syntactic processing: Figuring out the grammatical role of the word in the sentence (for example, is it a subject, a verb, an object?).
• Working memory use: Temporarily storing parts of the sentence while we continue reading.
• Long-term memory use: Bringing in knowledge from past experiences if needed to understand the text.

2. Problem: Reading should be much slower if these processes happened strictly one after the other

Each step takes time:

• Moving the eyes (making a saccade) takes about 125 to 200 milliseconds.
• Visual feature processing also takes measurable time (based on EEG experiments).
• Accessing the word's meaning (lexical access) takes about 100 to 150 milliseconds.
• Syntactic processing also requires time.
• In addition, just sending the visual information from the eyes to the brain (called "brain lag") takes about 15 milliseconds.

If you add up all the time needed for all these processes, it seems reading a single word should take much longer than what we actually observe.