Visual Attention & Visual Attention & Inhibition of ReturnInhibition of Return

Data Analysis SessionData Analysis Session

Dr. Jasper Robinson &Dr. Jasper Robinson &Dr. Jonathan StirkDr. Jonathan Stirk

Reminder of what we didReminder of what we did

►Before Easter we built a Posner Cueing Before Easter we built a Posner Cueing TaskTask

►Subjects had to detect the presence of a Subjects had to detect the presence of a target presented in one of 2 locations target presented in one of 2 locations (left or right box)(left or right box)

►Each trial presented a Each trial presented a peripheralperipheral pre-cue pre-cue (box outline highlighted) which was (box outline highlighted) which was non-non-predictivepredictive of the target location of the target location i.e. validity of the cue was 50:50i.e. validity of the cue was 50:50

What we expectWhat we expect

► Despite the fact that the pre-cue is of no real use, we Despite the fact that the pre-cue is of no real use, we cannot help but covertly move our attention to cannot help but covertly move our attention to where the peripheral cue iswhere the peripheral cue is

► This will lead to faster target detection times for This will lead to faster target detection times for VALID trials vs. INVALID onesVALID trials vs. INVALID ones BUT!!!!!BUT!!!!! If nothing happens at the cued location for a while then If nothing happens at the cued location for a while then

maybe attention will disengage and move back to the centremaybe attention will disengage and move back to the centre► Therefore we may expect the difference between Therefore we may expect the difference between

valid/invalid reaction times will be influenced by the valid/invalid reaction times will be influenced by the size of the delay between the cue and the target size of the delay between the cue and the target appearing appearing (CTOA )(CTOA )

A graph of possible dataA graph of possible data

► For short CTOAs we For short CTOAs we expect RTs for valid expect RTs for valid trials to be faster than trials to be faster than invalid trialsinvalid trials

► However, at some point However, at some point (around 200ms!) the (around 200ms!) the benefit for valid trials benefit for valid trials will reverse such that:will reverse such that:

► For longer CTOAs we For longer CTOAs we expect RTs for valid expect RTs for valid trials to be trials to be slowerslower than than invalid trialsinvalid trials

Black (filled circles) are valid trials

White are invalid trials

Structure/Timings of Exp’tStructure/Timings of Exp’t

CUE (Duration 100ms)

ISI/cue-target interval (Duration variable: 50, 100,

300)

TARGET (Duration 2000ms or until response)

TIME

The 2 key variables manipulated were CTOA and Validity of the cue

CTOA is the time from the ONSET of the Cue to the ONSET of the Target

CTOA of 150, 200 & 400 msecs

SUMMARY:

2 IVs

Validity - 2 levels

CTOA - 3 levels

1 DV – Reaction time (msecs)

Designs with more than 1 IV Designs with more than 1 IV (factor)(factor)

► These designs are more complicated than These designs are more complicated than simpler designs that manipulate only 1 IV at simpler designs that manipulate only 1 IV at 2 levels (like most of your practicals so far)2 levels (like most of your practicals so far)

► We would normally use a statistical We would normally use a statistical procedure known as procedure known as Factorial Analysis Of Factorial Analysis Of Variance (ANOVA)Variance (ANOVA) to examine the effects of to examine the effects of our IV manipulations on the dataour IV manipulations on the data

► However, as you have not been taught However, as you have not been taught Factorial ANOVA we are going to simplify the Factorial ANOVA we are going to simplify the design and use a design and use a ONE-WAY ANOVAONE-WAY ANOVA instead. instead.

Creating a simpler designCreating a simpler design

►We can collapse data over the Validity We can collapse data over the Validity factor by simply calculating a difference factor by simply calculating a difference scorescore We mentioned this the other week in the We mentioned this the other week in the

introductory sessionintroductory session We did a similar thing for the We did a similar thing for the

Pseudohomophone practical last semesterPseudohomophone practical last semester►Difference score = RT Invalid – RT ValidDifference score = RT Invalid – RT Valid►So now we can simply look at the effect So now we can simply look at the effect

of of CTOACTOA on these difference scores on these difference scores

Difference scoresDifference scores

► Difference score = Difference score = RT Invalid – RT ValidRT Invalid – RT Valid

► + ve diff = Faster + ve diff = Faster for valid cuesfor valid cues

► -ve diff = Slower for -ve diff = Slower for valid cuesvalid cues

► No difference (0) = No difference (0) = same for both same for both valid /invalid cues valid /invalid cues (cross over point)(cross over point)

Black (filled circles) are valid trials

White are invalid trials

Running a one-way ANOVA on Running a one-way ANOVA on the difference scoresthe difference scores

► From the data we calculate 3 difference From the data we calculate 3 difference scoresscores One for each level of our CTOA factorOne for each level of our CTOA factor

► If there is no effect of CTOA then the If there is no effect of CTOA then the difference scores won’t change (Null difference scores won’t change (Null hypothesis)hypothesis)

► If there is an effect of CTOA then we might If there is an effect of CTOA then we might expect to see some significant differences in expect to see some significant differences in the difference scores the difference scores i.e. a i.e. a main effectmain effect of CTOA of CTOA

What does a one-way ANOVA What does a one-way ANOVA tell us?tell us?

► It tells us whether any of the means It tells us whether any of the means (the difference score) (the difference score) significantlysignificantly differ from one another differ from one another

► It does It does NOTNOT tell us however, tell us however, whichwhich means differmeans differ

► If we find a significant main effect of a If we find a significant main effect of a factor (CTOA) then we need to follow-factor (CTOA) then we need to follow-up with what is called a up with what is called a post-hoc testpost-hoc test We can use We can use t-testst-tests to compare pairs of to compare pairs of

meansmeans

There are 2 types of one-way There are 2 types of one-way ANOVA!!!!ANOVA!!!!

►Which do we use?Which do we use? One-way within-subjects ANOVAOne-way within-subjects ANOVA One-way between-subjects ANOVAOne-way between-subjects ANOVA

►The answer depends on how we The answer depends on how we manipulated our factor (CTOA) in this manipulated our factor (CTOA) in this designdesign Did we test all subjects on all levels of Did we test all subjects on all levels of

CTOA or did we test 3 separate groups CTOA or did we test 3 separate groups each on one level of CTOA?each on one level of CTOA?

Lets run a within-subjects Lets run a within-subjects (related) one-way ANOVA(related) one-way ANOVA

► So lets get outputting your data and So lets get outputting your data and analysing itanalysing it We can follow-up the ANOVA with some related t-We can follow-up the ANOVA with some related t-

tests tests ifif there is a significant main effect there is a significant main effect

► So what data do we need and what form is it So what data do we need and what form is it currently in?currently in? We need mean difference scores to analyse for all We need mean difference scores to analyse for all

3 levels of our IV (CTOA)3 levels of our IV (CTOA) So we need 3 scores from each subjectSo we need 3 scores from each subject Currently we have RTs for valid and invalid trials in Currently we have RTs for valid and invalid trials in

separate .edat files (one for each subject)separate .edat files (one for each subject)

E-Prime data outputE-Prime data output

►You should each have tested 1 or 2 You should each have tested 1 or 2 subjectssubjects

►This produces .edat filesThis produces .edat files►Lets open one and examine what’s Lets open one and examine what’s

inside!inside!

You should have 144 rows of data for each subject

►Much of this Much of this output is output is not needed, not needed, so let’s just so let’s just look at what look at what we need:we need:

►ValidityValidity► ISIISI►Target.ACCTarget.ACC►Target.RTTarget.RT

►Right-click Right-click unwanted unwanted columns and columns and click Hideclick Hide

►ValidityValidity►ISIISI►Target.ACCTarget.ACC►Target.RTTarget.RT

►NoNo data has been data has been deleted, just deleted, just hidden to simplify hidden to simplify viewingviewing

So what do we need to So what do we need to analyse?analyse?

►We need to break the RT data up We need to break the RT data up according to Valid/Invalid trials and also according to Valid/Invalid trials and also the 3 ISI’s.the 3 ISI’s.

►We are only interested in Target Present We are only interested in Target Present trials so we can filter out trials so we can filter out notargnotarg trials trials using the using the validityvalidity column column This will reduce our trials by half!This will reduce our trials by half!

►We are also only interested in trials when We are also only interested in trials when you made a correct responseyou made a correct response So we can filter the data by Target.ACC tooSo we can filter the data by Target.ACC too

Filter out data from unwanted Filter out data from unwanted trialstrials

Filter out Target Absent trialsFilter out Target Absent trials

Choose the filter: Tools – Filter

Select Validity in the column name box

Then click ‘Checklist’

Tick only the valid and invalid trials (NOT Notarg!!)- Click OK

Your data will be visibly reduced by half

Let’s now get rid of Incorrect Let’s now get rid of Incorrect ResponsesResponses

Select Target.ACC in the column name box

Then click ‘Checklist’

Tick only the correct responses (1)

Your data may be reduced

So before we output the means So before we output the means make sure you have filtered the make sure you have filtered the

data!data!

Filters selected are listed here

We can now calculate the mean We can now calculate the mean RT’s for each cell of the designRT’s for each cell of the design

CTOA (Cue duration + ISI CTOA (Cue duration + ISI duration)duration)

150150 200200 400400

ValidityValidity

ValidValid Mean Mean RTRT

Mean Mean RTRT

Mean Mean RTRT

InvalidInvalid Mean Mean RTRT

Mean Mean RTRT

Mean Mean RTRT

Diff scrDiff scr Diff scrDiff scr Diff scrDiff scr

So we need to output the following means, which we can then copy into SPSS

We will however, calculate a difference score and then run the ANOVA

Let’s use the Analyze function in Let’s use the Analyze function in E-data aidE-data aid

►Bring Bring Validity and Validity and then ISI into then ISI into the columns the columns boxbox

►Subject into Subject into RowsRows

►Target.RT Target.RT into Datainto Data

►Click Click RunRun

This will give you your means This will give you your means for the 6 cells of the designfor the 6 cells of the design

► Now calculate Now calculate the difference the difference scores for scores for each of the 3 each of the 3 ISIsISIs

► Diff = Invalid-Diff = Invalid-ValidValid

► Write down 3 Write down 3 diff scores on diff scores on results sheetresults sheet

- = +55.35ms

For 50ms ISI

Record data for 2 Record data for 2 participantsparticipants

Let’s have a break whilst we Let’s have a break whilst we input your data into Excelinput your data into Excel

OK time to get data into SPSS OK time to get data into SPSS and analyse it!and analyse it!

►Really what we have manipulated is CTOAReally what we have manipulated is CTOA So you will need to So you will need to add 100add 100 (duration of cue) (duration of cue)

to each ISI to give the 3 levels of the CTOAto each ISI to give the 3 levels of the CTOA

►Set-up the variable names in SPSS (one Set-up the variable names in SPSS (one column for each CTOA)column for each CTOA)

►Run a repeated-measures ANOVARun a repeated-measures ANOVA►Follow up a signif. main effect with Follow up a signif. main effect with

Bonferroni Paired-t-testsBonferroni Paired-t-tests

Further info on SPSS can be Further info on SPSS can be found in:found in:

P 306 onwardsP 427

onwards