Your eyes are doing weird and wonderful things as you read these words. They flit along in distinct start-stop movements as your brain attempts to absorb the meaning behind the text.
The eyes fix every eight to 10 characters or so, pausing for about a quarter of a second to allow the mind to comprehend before jumping forward to the next eight. "These are ballistic movements," explained Dr Ronan Reilly, a cognitive scientist in the Department of Computer Science at UCD. "Your eye is effectively thrown to the next position."
Dr Reilly is leading a Forbairt- funded £42,000 two-year project to study eye movements during reading and create a computer model that can predict these movements. It got under way last Christmas and involves two student researchers.
The effort has an extra dimension of complexity in that models for three languages, English, Irish and German, are to be developed and then compared. "The goal here is to try to refine the model so it can handle variations in the three languages."
The language selection is deliberate in that the three represent very distinct sentence structure forms. The verb falls in the front of most Irish sentences, in the middle of English sentences and at the end of the sentence in German.
There are also significant differences in how much "extra" information can be derived from the grammar of a language, for example the use of word endings which can carry details about gender or ownership.
English has fewer inflections that carry meaning or embedded information, Dr Reilly said. "We don't get much information from the structure of the words," compared with some other languages.
The team is using a computer to analyse how the eyes and the brain link up to deliver language comprehension. "The main motivation in studying eye movement in reading is trying to understand how the brain understands language. My main interest is how people do language in their heads. The model is like a window into these processes."
The team is using 10 native speakers in the three languages who will all read the same three or four chapters of a novel. Their eye movements as they scan the texts on a computer screen will be established using an "eye tracker".
The tracker uses a tablemounted frame which holds the head in a fixed position during reading. An infra-red beam is pointed at the eye and a camera records the reflection coming back from the retina. This signal is interpreted by a computer, which can calculate a pinpoint location on the screen each time the eye pauses during the reading process.
The system must first be calibrated for each user by having them direct the eye towards designated points on the screen. Once this is done the person simply reads the text and the system collects all the raw data on eye movement.
There are obvious considerations when interpreting the data. The time spent looking at an eight-character block of text would "give you some information about how difficult the person is finding the language used". Familiar words are processed quickly but less familiar words and sentence structures are handled more slowly, forcing longer stops between eye movements.
There are also significant problems in interpretation, notably that the "eye is a bit decoupled from the brain", Dr Reilly explained. He likened this to breathing or walking, which are done without thinking unless the person decides to intervene in the process. The two processes here are eye movements to make the text available and language-processing by the brain.
"It is really necessary to use a computer model to understand these complex processes," he said. The model would allow him "to filter out the uninteresting bits and leave me with the interesting data" about how language is processed. "The big debate is what dictates how the eye moves." Does it seek out the verb, does it stop on the small or long words by preference, and how much understanding becomes available with each fix of the eye? Is it simply a matter of a fixed length step by step through a text, or can the text itself influence eye movement?
"The truth is somewhere in between," he believes. The business of reading is a very fast process, he added. "Contrasting the data from the three languages would help the debate" in understanding the process.
He will use the model to develop a baseline which will give good predictions about how the eye will respond to a given text. The team will then make subtle changes in the original text to see how the eye pattern is affected, which they hope will give them an insight into how the mind reaches comprehension.