Fixa%ons Microsaccades saccadic curvature Review: The visual coordinate system • Spaotopic coordinates • This is the world we think we see • Stable, unless objects are moving in the world • But, As we move our eyes, the image on the re$na changes • And neurons in the visual cortex are a 2D map of the re$na • Re$notopic coordinates • This is the world as it falls on the re$na of the eye • If objects in the world fall on different neurons, how does the visual system know they are the same object? Review: Re%notopic coordinates The nave frame of reference for human vision is re$notopic (Golomb, 2008) Review: Spaotopic coordinates ? This is the problem of Visual stability Maintained fixaon • We always say this is a period of ‘relave’ ocular stability Scleral coil. Ouch. • Earliest studies on small movements within fixaon used Scleral Eye Coil (or variant) • .01 dva • 1000+hz • Steinman, et al. (1973). Miniature eye Dual purkinje (corneal reflec$ons) movement. Science Front(1) and rear(2) reflec$ons of cornea • A lot is based on this early work Front (3) and rear(4) reflec$ons lens Fixa%onal Moon • Slow fixaonal oscilaons • dri], slow control • Occasional, small interrup$ons • Microsaccades • 2-3 per second • 12’ – 15’ • Usually binocular Slow control: Moon • If the visual cortex is a re$nal map of visual input • Then eye mo$on will create input ‘blur’ • Too much mo$on impairs vision (Burr&Ross, 1982) • High velocity mo$on from saccades (up to 900 Corollary Discharge. A corollary deg/sec) is blocked perceptually discharge (CD) is a copy of a motor • Saccadic suppression, aka saccadic masking command that is sent to the muscles to • Look in the mirror, you cannot see your own produce a movement. This copy saccades or corollary does not produce • Image mo$on caused by head movements any movement itself but instead is have less impact than mo$on from moving the directed to other regions of the brain to smuli inform them of the • Corollary discharge? impending movement. Slow control: Moon • Too liele mo$on is also a problem • Some mo$on is actually required for vision • Shi]ing the image to match eye mo$on causes the image to fade (collowyn & Kowler, 2008) • Remember gaze con$ngent? • Paralyzing eye muscles for cause temporary blindness • Increasing image mo$on increases acuity • But only up to a point • Acuity increases with mo$on, but only up to 2deg/sec • Mo$on is more beneficial for low spaal frequency Low sf High sf Microsaccades, fixaonal saccades and dri • Small saccades happen frequently during maintained fixaon • 2-3 per second • Some tasks as few as 1 every few seconds • Microsaccades previously classified at up to 15 arc- minutes (1/4 degree) • More recent allow up to .5+ (Engbert&Kleigl, 2003) • Microsaccade ‘creep’ due to • headrests vs bitebars, increased head mo$on • poor illuminaon control, • Higher noise with P+CR Microsaccade con%nuum Euclidean plot Velocity plot • Typically 12 – 15 arcminutes • But fixaonal saccades can exceed .5 degree • Just like larger saccades, there is a correlaon between velocity and amplitude Neural causes of microsaccades • Large saccades are mapped in vector space in the rostral superior colliculus • 2d map of all angles and amplitudes • Un$l fairly recently we thought the deepest rostral area was fixaon maintenance and saccadic inhibi$on • But… microsaccades are on neural con$nuum with full saccades. • Mapped in the deep rostral of SC : (Hafed, Goffart, and Krauzlis (2009)) • Best current theory model places ms completely in SC (Hafed, 2009) • Rostral SC neurons monitor average SC ac$vity as compared to a point of focused fixaon • Changes in mean ac$vity, if large enough, trigger a microsaccade • Mean ac$vity could change due to dri], object mo$on, Our very own Tanya head mo$on, simple neural noise, etc Malevich recently • SC ac$vaon is also influenced by top down aen$on joined Ziad Hafed’s (FEF), allowing for some cogni$ve control team in Tubingen for her PhD Microsaccade funconality • Neural con$nuum, but what about func$onality • Fixaonal saccades larger than 20’ have the same func$onality as larger saccades • Simply adjus$ng centre of gaze to compensate for decline in resolu$on • But smaller (classic < 15’) microsaccades serve a special func$on • Acuity? • No. Microsaccades were uncommon and do not improve performance on needle threading (Bridgeman & Palca, 1980) • But do occur looking back and forth between needle and thread (Ko etal, 2010) • Prevent fixaonal fading? • No. Fixaon dri] does this beeer • Any mo$on is good, but no fading even without microsaccdaes (Collowijn, 2010) • Aen$on? • Maybe. There is a connec$on here at least • Direc$on and frequency Microsaccades and aen%on • Modulated by aen$on (Hafed & Clark, 2002) • Changes for exo and endo (Laubrock et al, 2010) • Microsaccade inhibi$on at target onset (Engbert & Kleigl, 2003) • Top down modulaon to improve signal quality Counter aen%on • Horowitz et all (2007) disagree (to an extent). • Microsaccades are not an independent measure of aen$on • When microsaccade direc$on and the cue locaon disagree, the cue locaon predicts aen$onal benefits, not ms direcon • They agree that most o]en cue, ms and aen$onal benefits agree • But they argue that’s because the Cue causes both ms direc$on and aen$on • When they differ, it’s the cue that wins • MacInnes and Bhatnagar (2018, Scien$fic Reports) agree with Horowitz • In a paradigm that does not produce robust aen$onal facilitaon • There is also minimal impact on ms rates nor any impact on direcon MacInnes and Bhatnagar • The debate is ongoing (Laubrock, Kliegl, Rolfs, & Engbert, 2010). (2018, Scien$fic Reports) Other uses • Fixaonal saccades can revive extrafoveal (3-9 degrees) fading • Where slow control might be insufficient mo$on (Mar$nez-Conde, 2006) • Hafed (2009) suggested microsaccades were correc$ons to fixaon locaon • Error caused by dri], etc • Triggered only by mean SC neural ac$vity level Microsaccades, fixaonal saccades and slow control • Microsaccades are not the primary means of maintaining stable fixaon. • Micro Saccade rates can be reduced by simple voluntary effort • stable fixaon maintained exclusively by slow eye movements – slow control. • Slow control acts to maintain stable fixaon mainly by controlling the velocity of the re$nal image, rather than by correc$ng offset errors in fixaon posi$on (Epelboim& Kowler, 1993). • Microsaccades • do not possess the characteris$cs of an involuntary reflex, • nor a special class of eye movements, but rather have the same basic proper$es as their larger, unambiguously voli$onal, counterparts. Binocular Microsaccades Typically there is a high correlaon between eyes for microsaccade rate and direc$on Binocular microsaccades may be beeer indicators than the less frequent monocular Image source: Engbert 2006 Velocity threshold algorithm: (Engbert & Kliegl, 2003) Velocity based detec$on, similar to larger saccades Step 1: The $me series of eye posi$ons is transformed to veloci$es, V(n) = x(n+2) + x(n+1) – x(n-1) – x(n-2) 6 Δt which represents a moving average of veloci$es over 5 data samples. Step 2: Compute velocity thresholds for horizontal and ver$cal components (based on a mul$ple of S.D of velocity distribu$on). Step 3: Define dis$nc$on between monocular and binocular microsaccades Step 4 (op$onal): Verify algorithm validity by plong ‘main sequence’. Main sequence Toolbox available at - hep://read.psych.uni-potsdam.de/index.php? op$on=com_content&view=ar$cle&id=140:engbert-et-al-2015-microsaccade-toolbox-for- r&cad=26:publicaons&Itemid=34 Image source: Engbert & Kliegl, 2003 Unsupervised clustering: (Otero-Milan et al., 2014) Step 1: The $me series of eye posi$ons is transformed to veloci$es, Vi = Fs (x(i+2) + x(i+1) – x(i-1) – x(i-2) ) 6 where xi is the eye posi$on (horizontal or ver$cal) at $me i, vi is the instantaneous eye velocity (horizontal or ver$cal) at $me i, and FS is the sampling rate. Step 2: Calculate velocity magnitude from horizontal & ver$cal components. Step 3: Use k-means to arrive at two poten$al clusters – microsaccades and noise. Features considered are peak velocity, ini$al acceleraon, final acceleraon. MATLAB implementaon at - hep://smc.neuralcorrelate.com/sw/microsaccade-detec$on/ Interlude: Smooth pursuit • O]-forgoeen eye movement type • Just one slide here since they aren’t relevant to most of our research • Smooth tracking is possible only with external s$mulus • We can combine saccades with smooth pursuit as we track mul$ple objects • New research on ‘predic$ve’ tracking • Top down control of tracking can take advantage of scene knowledge Saccadic Trajectory Saccades do not travel directly from fixaon to target May be different for horizontal and ver$cal saccades And also across individuals Van Der Sgchel, 2010 Review Curvature direc$on • We can calculate the direc$on of any saccade curvature - • Absolute coordinates: Up, down, le] right target • More interes$ng is whether the curvature is influenced by some distractor, onset, cue, + etc… onset • This relave direc$on can be coded as toward or away from the relave event • Toward and away may be influenced by many factors Influence on curvature Toward t2 • Double step paradigms (McPeek, 2000) • Instruc$on of Saccade to target, but aer a variable delay, it’s replaced by a new t1 target • If the delay between onset of first and second target is short (50 ms) saccades will curve through first target on the way to second Influence on curvature Toward • Global effect (Walker eta al, 1997) • Landing posi$on will typically be t between target and adjacent irrelevant distractor d • Curvature