Eye Tracking in Reading: Experimental Design and Cognitive Insights

Hello and welcome to the course basics of experimental design for cognitive psychology. I am Dr. Arkwarma from the

department of cognitive science at ID Kpur. This is the final lecture of week seven and we are discussing eye tracking

as a methodology. Now in today's lecture I will talk about eyetracking from a perspective of how to

analyze reading. Now uh before we go with the eyetracking uh sort of uh things uh let's first understand what

reading is. Now reading as uh you know indicated by go and hilling uh is an unnatural act. It is not something that

we are born with. It is not something that uh you know language evolved and say for example it's not an innate

capability. It is something uh that everybody has to learn. We have to learn to navigate with the uh graphims. We

have to learn with navigates with navigate with the letters. We have to learn to connect letters. We have to

learn to sort of make sounds from letters and then connect those sounds to meaning. So in that sense reading is

actually an unnatural act that involves close coordination of motor, visual and cognitive functions. How does let us let

us revisit it again and again. uh what is motor say for example when you're reading a particular sentence uh or even

a given word you are actually scanning through the making of that letter all right so you're scanning through the

making of a letter then another letter then another letter and then you're sort of going in the direction of how the

script is written say for example for Hindi and English it is left to right so you are scanning from the first letter

and then you go to the second letter third letter then you go to the next word left to right then you go to the

next word then you go to the next word And that is basically how you are doing. So you are moving your eyes. You are

basically uh you know planning the movement of your eyes in such a way that it is getting you most information out

of the text. All right. So uh first word in the phobia remember one to two degrees that is the where you are

getting most information out of. So first word then you move to the second word get most information out of it.

Then you move to the third word then you get most information out of it. So you're motor you're basically planning

the movement of your eyes in such a uh manner and when you're moving your eyes what is it that you're doing you're

getting visual information out of the text you're basically getting the details of how a letter is written. Say

for example there is a small difference between a b and a p if you are an English reader. There's a small

difference between a p and a p if you are you know hindi reader. So you're basically what you're trying to do is

you are trying to also unlock what is the letter that is written and you have to attend to it in its extreme detail

because uh you know letter say for example especially if you talk about handwriting different people write

things differently. Print has certainly sort of created a degree of homogeneity uh as to say for example given a font

this is how a letter letter will be written. So you're moving your eyes to get maximum information and what is the

information that you're getting? You're getting the information of what was that letter that is written there. What is

the sound that that letter represents and how do you connect the letters and the sounds that they make to make a word

and then you basically consult your mental lexicon for the meaning of that word and that is basically how you read.

So you read by making motor movements to sample visual information to get cognitive details about what is written

in the text. That is what reading is all about. So it is motor, visual and cognitive coordination basically

facilitated by the movements of eyes over text. So that is why reading is a very interesting case where eye

movements have been found especially useful to get maximum information out of. Now uh what kind of cognitive

processes uh you know uh are involved in reading? Uh can you think of what we do say for example in order to be able to

read? As I just said you're basically sampling the visual information. You are sampling the visual patterns. You're

getting information out of it. You're uh uh deciphering those patterns to unlock the sounds that those patterns

represent. You are concatenating not only those letters as they are written but also the sounds that they are

making. And when you're concatenating those sounds, you're getting to what are words. So for example, uh mutkar gpar

matal something like that. You are basically collecting those sounds. Those sounds mean something. you get the

meaning of that and that is how you read. You're not only understanding the meaning of an individual word but you

are basically integrating that meaning of that individual word in the sentence. You are integrating the meaning of that

sentence in the overall discourse and that is basically how you are uh executing what is reading. So in that

sense you can again see that movement of eyes is able to provide you with a lot of information and it is able to

facilitate the process of deriving information from text and reading and understanding whatever uh material is

written there. So to read efficiently people basically need to extract visual information from the environment and

this has to be done relatively rapidly because our reading speeds are extremely good. they need to activate stored

phonological that is sound based and semantic that is meaning based representations in the right way in the

right order and at the right time. So reading is basically quick scanning of text to uh get to sound and get to

meaning in the right order at the right speed so that you can unpack the text while you're reading at a particular

pace. Also while you're reading you are also understanding it. You are understanding

not only the word, you're understanding not only the sentence, but you're also understanding the overall text, you're

understanding what is this story about, what is this uh you know, let's say you're reading poetry, what is this

poetry about. So higher cognitive processes are also engaged while you are actively sampling the visual space to

understand what these letters are, what the sound they represent, what kind of uh meaning that is uh you know uh

emerging from there. So actively we have to uh you know on the fly activate stored phonological representations and

semantic representations in the right way at the right order at the right time. So you're doing it rather rapidly

and also engaging in simultaneously in higher cognitive processes and inferential processes to understand what

is the meaning that the text contains. Also uh while you're reading text you have to plan your eye movements in a

particular way. So I'm reading this sentence and I also have to plan okay next I will go here the next I will go

here and then I'll come down and then I'll move again left to right then I'll again come down then I'll again move

left to right. So bunch of processes are happening at the same time while your eyes are moving across text and sampling

information out of it. Now when you talk about this how do you sort of figure this out? There are two things. First is

how do we control our eye movements during reading? What are the higher cognitive aspects that are involved in

reading? And you can ask several questions here. For example, how the different kinds of scripts affect the

way we extract information? How do children learn to read this uh script? Why do some children and adults have

difficulty reading? Some of these questions can actually and are actively uh sampled through eyetracking research.

Uh some of which I will just give you a very small uh you know uh brief into and then uh at later point we will talk

about them. Now eye movements in reading. So you you have a page, you have a piece of text. Nowadays we mostly

have screens uh whether it is a cell phone screen or a laptop screen or a tablet screen we look at this and you

are scanning our eyes over it. Are we reading? U not really unless you are actually understanding. Just scanning

plainly does not mean reading. Uh you have to scan while getting information out of the screen. That is when you're

reading. So eye movements in reading if you see are systematically organized. They're not random eye movements.

They're not random eye movements over the screen all over the place. All right. So, eye movements in reading are

systematically organized so that one starts in one location on a page and then ends up in another location. So,

you start from the top left of the page, then you move to the uh you know from left to right. You need uh read uh the

end of the sentence and you come back again in the beginning uh and the left side then you again go from left to

right then you come come again on the left of the uh page and so on. Okay. So that is basically how planned and

systematic your eye movements are. End product of this behavior basically helps us formulate mental representations of

whatever is the text that we are reading. Two broad kinds of eye movements uh some

of which we've already studied. So there is one smooth pursuit when you are following something. So for example you

uh you know uh standing somewhere and you're passing you following the cars that are passing by. So you are smoothly

following the movement of the cars. That is called smooth pursuit. But the more important ones for reading that we also

discussed in the previous lecture are psychic eye movements. Psychadic eye movements occur when you're reading text

and you're scanning written material in the direction of the text. So for example, if you're reading Hindi, you're

going from left to right. If you're reading udu, you're going from right to left. So psychic eye movements in the

direction of the text are informative about whatever text is written. When there are relatively longer periods of

time when the eyes are stationary in their orbits that is fixation and these secs fixation sad fixation sad that is

basically connected by short bursts of very rapid movements that are called secads. Sec around 20 milliseconds from

start to finish. So once the eye takes off it takes around 20 millconds to land at some place again takes off takes

around 20 millconds to land at some place. The purpose of these secads is to sequentially bring each part of the text

in the fobial vision because if you remember I told you phobia is the region that provides highest visual acquity

that that has the highest concentration of cones that provides the highest visual equity information from the text.

Why is visual equity important? It is important because you cannot only you can only tell the difference between

different letters by attending to extremely small details in those letters. Whether it is whichever script

you're reading, you're reading Hindi, Tamil, Turbo, Bangla or English or French or Dutch. Uh there are small

details that you need to attend to. Say for example N versus M. Very small detail. There's one extra curve there. P

versus Q the direction of the semi close semicircle is just uh on the other side in P versus Q. B versus P when you are

writing in small case. So these are uh the things that you have to pay attention to if you're reading rapidly

and you see people read very well rapidly and they don't make a lot of errors because your eyes are adept at

moving at certain speeds scanning the text getting you the best details out of the text and that is because uh you are

moving your eyes to fiate on the specific words and get the highest resolution accurate information out of

the text. Now rapid eye movements basically when you make a sakad from point A to point B is linked with what

is called psychadic suppression. Secchadic suppression is the period when the eyes are a flight then there is no

uh you know uh information that can be extracted while the psychicad is in progress. Partly it is because the

visual image on the retina basically is consisting of a big blur because your eyes are moving. They're not resting and

gaining information. also because during that period of the eye your eyes is actually in motion. So there is very

little time uh for the activity at the retina to uh you know uh uh activate uh corresponding areas in the visual cortex

and new information cannot be analyzed as quickly. Now recent research basically shows that

at least 50 millisecond of stable exposure is necessary for us to be able to uh identify words or extract enough

useful visual information. So this less than 50 milliseconds is called the eye brain lag. That is the amount of time it

takes for the neural signals at the retina to reach the neural signals in the primary visual cortex that is the

area V1 at the back of the brain to get information. So the eyebrain lag is around 50 milliseconds. 50 milliseconds

the eye must rest at some place for you to be able to process that particular site

and there are uh pertaining to reading there are different kinds of psychic eye movements there are progressive sads

which are in the direction of the text so when you're reading Hindi or English left to right uh that is a progressive

sakad if the direction of the text is left to right and the eyes come back at some points that is called a regressive

sakad against the direction of the script okay so eye movements from left to Right? While reading languages like

Hindi etc are called progressive sakads. Sakads that move back on the text that is right to left when the text is

written left to right is called a regressive sakad. Why does a regressive sakad happen? Remember we were talking

about it. Regression usually happens when something has gone wrong when something you've not understood. Say for

example he spread butter with warm socks. Something like that. If you read uh some sentence like that or you spread

the warm bread with uh you know instead of jam with oil or something like that. Now when you are not able to integrate

the meaning of a word in uh overall sense uh you might want to revisit and you might want to say oh did I read it

correctly? He spread the uh you know bread with socks. Something like that. Now one does not usually spread the

bread with socks. So you might want to come back and read oh is it actually what I read. So regression usually will

happen when something has gone wrong with the process involved in interpreting and comprehending the

written text. Okay. Regressions are targeted towards part of the text that may be helpful in resolving and uh these

comprehension problems. Say for example something that is not understood. Remember uh that garden pass sentence a

horse race past the barn fell. Now the eyes keep coming back to the barn. They want to understand what is it that has

happened here? How do I resolve this uh you know ambiguity? So the readers typically they maintain a spatial map to

keep track of syntactic choice points the points at which important information is unfolded and other

potentially ambiguous material. So where say for example you know ga ga ga so where is it you know that the same word

acts as a noun where does it act as a verb where does it act as an object you want to read it and you want to unpack

what is it that I did I you know that I have read fixations especially in terms of reading

when skill adults read they fixate directly on the words on the page so it takes some time for uh you know children

to pick up the skill to fixate directly on the exact act word that you have to read on the page gaze will directly land

on some part of the word and stays there for around 250 to 500 milliseconds on an average to get the full information from

that word some words are however often skipped owing to their predictability say for example when you're reading

English uh articles prepositions etc are highly predictable and those are some of the shorter words that you will see that

they get skipped while a person is reading you intake information from uh of up to a whole line of text when

you're reading into a single fixation. So when your eyes are landed here, you probably are getting information from

the whole line of text at the same time from adjacent lines sometimes and even from the lines above and below. That is

basically this region from which you are getting maximum information out of is referred to as the perceptual span. So

perceptual span which is the region of useful vision uh from where the skilled readers draws the maximum information

out of is called the perceptual span for uh when the text is left to right it is four characters uh uh to the left and 15

characters to the right. When the writing is right to left it is four characters to the right and 15

characters to the left. So basically this entire region from which you will get maximum information out of

this is basically diagram of reading. So for example, roadside joggers endure sweat, pain and angry drivers in the

name of fitness. So see eyes landed here uh just around middle of that word and then when two joggers endure sweat they

skipped pain and pain and then go to angry drivers in uh the name the is skipped the name of fitness. Uh a

healthy body may seem reward. Now here something is sort of a miss here. So the person goes back there's a regressive

sakard on healthy again and then it goes forward again. Uh and here you can see basically uh you know the equity this is

where the phobia is and this is how the equity map changes. So just next to the phobia there is 75% equity and the

equity keeps uh decreasing as you move away from the fobial position. So from only four to five letters uh which are

seen which have the maximum equity that is why we need to move our eyes in a certain way all the time. This is

basically the description. So 1 to 2° is the phobia. That is the maximum best resolution that you will get. 2 to 5° is

the paraphobia and after 5° is your paraphobia is your peripheral vision from 5° to around 200° that we saw in

the last class. This is again another graphic of uh how this is. So paraphobia peripheral

vision, paraphobial vision, near peripheral vision and then fobial vision of around 2°.

Now uh for reading people have devised different kinds of uh you know uh experimental methods. This is what I

want to talk about uh to study reading. So one is referred to as the moving window paradigm. Here part of text is

placed by something else. Let's say for example that is uh you know not giving you any information to manipulate the

part of text that is available to the readers while they are fixating. So this can be done very easily using feedback

from an eye tracker. An eye tracker basically can determine where these people are looking. We you know that

when only one word is visible at a time and the word boundaries are basically all X's and the reading speeds basically

become 20% slow. So you can look at this give me three steps. So here you can see uh it is only the word that is fixating

that is revealed everything to the left and the right of the word is just covered in excess and this is called a

moving window paradigm because this is basically where uh you know the eyes are landing and you're basically uh denying

the uh paraphobial uh information to the eyes for uh decoding anything. So no anticipation etc can happen here. There

is another thing which is called a boundary change experiment. It is a similar one. You can see here uh give me

and then there is a boundary everything else to the right of the boundary you can see is all uh uh you know just

jumbled letters and does not provide any information. So when the reader's gaze is left of the boundary the critical

text is either displays as normal or it is distorted in some way to the right everything is distorted in jumbled

letters. The purpose basically is is there to deny the readers any preview of the next letter. So you basically want

that the reader is getting only information out of the point from where he's fixating the most.

Some recent research suggests that readers can also sometimes identify the final letter in the word to the right if

the word is five or fewer letters wrong. So for example remember the phobia is around 2 to 5° only. All right. So when

your eyes are landing on a word uh and if the word is long sometimes it is possible that the last letter is not

visible. But say for example in some cases it is possible that the last letter is also sort of there but it is

slightly blurriier and there is this visual interference from adjacent letters that is called lateral masking

that is uh you know uh reducing the equity in that whole letter. Being able to identify letters from the word to the

right of the fixation to this side helps reader prepare to process the you know that a word uh that word at a later

point in time possibly by increasing the activation of phonological words. Basically when you're reading long words

you want to have the full information of the word and you want to sometimes also have information about the adjacent

words so that you can plan oh whether I have to read that word or whether I can skip and move to the next word. So that

is something that is important but uh at this point semantic information is not available semantic activation is not

happening uh at least uh you know and if the preview is there uh preview is not there then you cannot get that

information and anticipation is not there. So the reading basically will get slowed if the preview is denied. There

are models of reading the easy reader model of uh reading the swift uh you know model of reading that will uh

basically uh provide some more details about how does the process of reading work in such cases.

So size of the reading window does matter. The perceptual span, four letters to the left, 15 letters to the

right or when you're reading Hebrew or Udu, four letters to the right and 15 letters to the left. That basically uh

matters because that provides you with all the earlier uh anticipatory information both about phology and both

about semantics so that you can decide how fast you're going to be able to read, how easy it is going to be predict

the next uh you know uh elements of text and so on. So a onelet window uh basically when you sort of uh reduce

this window a onelet uh window led to the slowest reading times and uh basically people's reading time

increased several times uh you know uh when the reading window was actually reduced. So and also uh as you just

notice the perceptual span is not symmetrical. It is not four to left four to the right. It is basically four to

the left and 15 to the forward right. so that uh you can actually anticipate how information from text will be gained. So

what is the best site for getting reading information that is the optimal viewing position and when people are

reading a word that is five or seven characters long the entire word will fall within the phobia. But if there are

longer words basically people will fixate uh just left to the middle of the word and that is basically what has been

found in research to be the optimal viewing position. Just left. Say for example if it is a fiveletter word just

left to the middle of the word is the optimal position that allows you to max sample the maximum region of the word.

Okay. So for longer words the some part of the word may fall on the paraphobia let's say 6° from the from the uh center

of the vision and sometimes these kinds of words may take two fixations for us to be able to fully and properly read

them. The optimal viewing position basically provides us word processing system with

a stimulus that uh enables the fastest uptake of information to identify the word and that is basically what uh you

know enables faster reading of text. So eye movement planning basically uh expert eye movement planning is able to

pick up the optimal viewing position in the next word as it is uh you know planning the uh next sad when it is

fixated here it is already planning where do I land my eyes next so that I can reach here quickly and then I can

reach the next word quickly and so on. If the psychiatrist successfully lands on the optimal viewing position of the

next word, lexical access procedures start and they have the best possible visual image to get maximum information

out of. When word length options are degraded such as in the moving window uh kind of paradigm, the sadic planning

mechanism will not be able to identify what is uh landing next and the target optimal viewing position and the readers

will not have the ideal view scenario. So therefore the reading speeds will come down. Okay. Finally, fiating.

Basically, why do we fiate? We fiate to get the best vision out of the letters. Uh we want the uh most informative part

of the letter to fall within the phobia so that we get the best information out of uh that and that is basically why we

are moving our eyes a lot because we are trying to reorient our gaze so that the center of the object center of the word

falls in the center of the phobia. So that is basically uh most uh about reading that I wanted to talk about with

respect to eyetracking. I'll continue talking about other methods in the next lecture. Thank you.