Roy S. Hessels

What to expect from your remote eye-tracker when participants are unrestrained

The marketing materials of remote eye-trackers suggest that data quality is invariant to the position and orientation of the participant as long as the eyes of the participant are within the eye-tracker’s headbox, the area where tracking is possible. As such, remote eye-trackers are marketed as allowing the reliable recording of gaze from participant groups that cannot be restrained, such as infants, schoolchildren and patients with muscular or brain disorders. Practical experience and previous research, however, tells us that eye-tracking data quality, e.g. the accuracy of the recorded gaze position and the amount of data loss, deteriorates (compared to well-trained participants in chinrests) when the participant is unrestrained and assumes a non-optimal pose in front of the eye-tracker. How then can researchers working with unrestrained participants choose an eye-tracker? Here we investigated the performance of five popular remote eye-trackers from EyeTribe, SMI, SR Research, and Tobii in a series of tasks where participants took on non-optimal poses. We report that the tested systems varied in the amount of data loss and systematic offsets observed during our tasks. The EyeLink and EyeTribe in particular had large problems. Furthermore, the Tobii eye-trackers reported data for two eyes when only one eye was visible to the eye-tracker. This study provides practical insight into how popular remote eye-trackers perform when recording from unrestrained participants. It furthermore provides a testing method for evaluating whether a tracker is suitable for studying a certain target population, and that manufacturers can use during the development of new eye-trackers.

Noise-robust fixation detection in eye movement data: Identification by two-means clustering (I2MC).

Eye-tracking research in infants and older children has gained a lot of momentum over the last decades. Although eye-tracking research in these participant groups has become easier with the advance of the remote eye-tracker, this often comes at the cost of poorer data quality than in research with well-trained adults (Hessels, Andersson, Hooge, Nyström, & Kemner Infancy, 20, 601–633, 2015; Wass, Forssman, & Leppänen Infancy, 19, 427–460, 2014). Current fixation detection algorithms are not built for data from infants and young children. As a result, some researchers have even turned to hand correction of fixation detections (Saez de Urabain, Johnson, & Smith Behavior Research Methods, 47, 53–72, 2015). Here we introduce a fixation detection algorithm—identification by two-means clustering (I2MC)—built specifically for data across a wide range of noise levels and when periods of data loss may occur. We evaluated the I2MC algorithm against seven state-of-the-art event detection algorithms, and report that the I2MC algorithm’s output is the most robust to high noise and data loss levels. The algorithm is automatic, works offline, and is suitable for eye-tracking data recorded with remote or tower-mounted eye-trackers using static stimuli. In addition to application of the I2MC algorithm in eye-tracking research with infants, school children, and certain patient groups, the I2MC algorithm also may be useful when the noise and data loss levels are markedly different between trials, participants, or time points (e.g., longitudinal research).

The area-of-interest problem in eyetracking research: A noise-robust solution for face and sparse stimuli

A problem in eyetracking research is choosing areas of interest (AOIs): Researchers in the same field often use widely varying AOIs for similar stimuli, making cross-study comparisons difficult or even impossible. Subjective choices while choosing AOIs cause differences in AOI shape, size, and location. On the other hand, not many guidelines for constructing AOIs, or comparisons between AOI-production methods, are available. In the present study, we addressed this gap by comparing AOI-production methods in face stimuli, using data collected with infants and adults (with autism spectrum disorder [ASD] and matched controls). Specifically, we report that the attention-attracting and attention-maintaining capacities of AOIs differ between AOI-production methods, and that this matters for statistical comparisons in one of three groups investigated (the ASD group). In addition, we investigated the relation between AOI size and an AOI’s attention-attracting and attention-maintaining capacities, as well as the consequences for statistical analyses, and report that adopting large AOIs solves the problem of statistical differences between the AOI methods. Finally, we tested AOI-production methods for their robustness to noise, and report that large AOIs—using the Voronoi tessellation method or the limited-radius Voronoi tessellation method with large radii—are most robust to noise. We conclude that large AOIs are a noise-robust solution in face stimuli and, when implemented using the Voronoi method, are the most objective of the researcher-defined AOIs. Adopting Voronoi AOIs in face-scanning research should allow better between-group and cross-study comparisons.

Qualitative tests of remote eyetracker recovery and performance during head rotation

What are the decision criteria for choosing an eyetracker? Often the choice is based on specifications by the manufacturer of the validity (accuracy) and reliability (precision) of measurements that can be achieved using a particular eyetracker. These specifications are mostly achieved under optimal conditions—for example, by using an artificial eye or trained participants fixed in a chinrest. Research, however, does not always take place in optimal conditions: For instance, when investigating eye movements in infants, school children, and patient groups with disorders such as attention-deficit hyperactivity disorder, it is practically impossible to restrict movement. We modeled movements often seen in infant research in two behaviors: (1) looking away from and back to the screen, to investigate eyetracker recovery, and (2) head orientations, to investigate eyetracker performance with nonoptimal orientations of the eyes. We investigated how eight eyetracking setups by three manufacturers (SMI, Tobii, and LC Technologies) coped with these modeled behaviors in adults. We report that the tested SMI eyetrackers dropped in sampling frequency when the eyes were not visible to the eyetracker, whereas the other systems did not, and discuss the potential consequences thereof. Furthermore, we report that the tested eyetrackers varied in their rates of data loss and systematic offsets during shifted head orientations. We conclude that (prospective) eye-movement researchers who cannot restrict movement or nonoptimal head orientations in their participants might benefit from testing their eyetracker in nonoptimal conditions. Additionally, researchers should be aware of the data loss and inaccuracies that might result from nonoptimal head orientations.

Is There a Limit to the Superiority of Individuals with ASD in Visual Search

Superiority in visual search for individuals diagnosed with autism spectrum disorder (ASD) is a well-reported finding. We administered two visual search tasks to individuals with ASD and matched controls. One showed no difference between the groups, and one did show the expected superior performance for individuals with ASD. These results offer an explanation, formulated in terms of load theory. We suggest that there is a limit to the superiority in visual search for individuals with ASD, related to the perceptual load of the stimuli. When perceptual load becomes so high that no additional task-(ir)relevant information can be processed, performance will be based on single stimulus identification, in which no differences between individuals with ASD and controls have been demonstrated.

An in-depth look at saccadic search in infancy.

Two questions were posed in the present study: (1) Do infants search for discrepant items in the absence of instructions? We outline where previous research has been inconclusive in answering this question. (2) In what manner do infants search, and what are the fixation and saccade characteristics in saccadic search? A thorough characterization of saccadic search in infancy is of great importance as a reference for future eye-movement studies in infancy. We presented 10-month-old infants with 24 visual search displays in two separate sessions within two weeks. We report that infant saccadic search performance at 10 months is above what may be expected by our model of chance, and is dependent on the specific target. Infant fixation and saccade characteristics show similarities to adult fixation and saccade characteristics in saccadic search. All findings were highly consistent across two separate sessions on the group level. An examination of the reliability of saccadic search revealed that test-retest reliability for oculomotor characteristics was high, particularly for fixation duration. We suggest that future research into saccadic search in infancy adopt the presented model of chance as a baseline against which to compare search performance. Researchers investigating both the typical and atypical development of visual search may benefit from the presented results.

Interocular conflict attracts attention

During binocular rivalry, perception alternates.between dissimilar images presented dichoptically. Since.its discovery, researchers have debated whether the phenomenon is subject to attentional control. While it is now clear that attentional control over binocular rivalry is possible, the opposite is less evident: Is interocular conflict (i.e., the situation leading to binocular rivalry) able to attract attention?In order to answer this question, we used a change blindness paradigm in which observers looked for salient changes in two alternating frames depicting natural scenes. Each frame contained two images: one for the left and one for the right eye. Changes occurring in a single image (monocular) were detected faster than those occurring in both images (binocular). In addition,monocular change detection was also faster than detection in fused versions of the changed and unchanged regions. These results show that interocular conflict is capable of attracting attention, since it guides visual attention toward salient changes that otherwise would remain unnoticed for longer. The results of a second experiment indicated that interocular conflict attracts attention during the first phase of presentation, a phase during which the stimulus is abnormally fused [added].

Is human classification by experienced untrained observers a gold standard in fixation detection

Manual classification is still a common method to evaluate event detection algorithms. The procedure is often as follows: Two or three human coders and the algorithm classify a significant quantity of data. In the gold standard approach, deviations from the human classifications are considered to be due to mistakes of the algorithm. However, little is known about human classification in eye tracking. To what extent do the classifications from a larger group of human coders agree? Twelve experienced but untrained human coders classified fixations in 6 min of adult and infant eye-tracking data. When using the sample-based Cohen’s kappa, the classifications of the humans agreed near perfectly. However, we found substantial differences between the classifications when we examined fixation duration and number of fixations. We hypothesized that the human coders applied different (implicit) thresholds and selection rules. Indeed, when spatially close fixations were merged, most of the classification differences disappeared. On the basis of the nature of these intercoder differences, we concluded that fixation classification by experienced untrained human coders is not a gold standard. To bridge the gap between agreement measures (e.g., Cohen’s kappa) and eye movement parameters (fixation duration, number of fixations), we suggest the use of the event-based F1 score and two new measures: the relative timing offset (RTO) and the relative timing deviation (RTD).

The disengagement of visual attention in the gap paradigm across adolescence

Attentional disengagement is important for successful interaction with our environment. The efficiency of attentional disengagement is commonly assessed using the gap paradigm. There is, however, a sharp contrast between the number of studies applying the gap paradigm to clinical populations and the knowledge about the underlying developmental trajectory of the gap effect. The aim of the present study was, therefore, to investigate attentional disengagement in a group of children aged 9–15. Besides the typically deployed gap and the overlap conditions, we also added a baseline condition in which the fixation point was removed at the moment that the target appeared. This allowed us to reveal the appropriate experimental conditions to unravel possible developmental differences. Correlational analyses showed that the size of the gap effect became smaller with increasing age, but only for the difference between the gap and the overlap conditions. This shows that there is a gradual increase in the capacity to disengage visual attention with increasing age, but that this effect only becomes apparent when the gap and the overlap conditions are compared. The gradual decrease of the gap effect with increasing age provides additional evidence that the attentional system becomes more efficient with increasing age and that this is a gradual process.

Gaze Behavior to Faces during Dyadic Interaction

A long-standing hypothesis is that humans have a bias for fixating the eye region in the faces of others. Most studies have tested this hypothesis with static images or videos of faces, yet recent studies suggest that the use of such “nonresponsive” stimuli might overlook an influence of social context. The present study addressed whether the bias for fixating the eye region in faces would persist in a situation that allowed for social interaction. In Experiment 1, we demonstrate a setup in which a duo could engage in social interaction while their eye movements were recorded. Here, we show that there is a bias for fixating the eye region of a partner that is physically present. Moreover, we report that the time 1 partner in a duo spends looking at the eyes is a good predictor of how long the other partner looks at the eyes. In Experiment 2, we investigate whether participants attune to the level of eye contact instigated by a partner by having a confederate pose as one of the partners. The confederate was subsequently instructed to either fixate the eyes of the observer or scan the entire face. Gaze behaviour of the confederate did not affect gaze behaviour of the observers. We conclude that there is a bias to fixate the eyes when partners can engage in social interaction. In addition, the amount of time spent looking at the eyes is duo-dependent, but not easily manipulated by instructing the gaze behaviour of 1 partner. Une hypothèse de longue date veut que les humains aient tendance à fixer la région des yeux sur le visage d’autres personnes. La plupart des études ont testé cette hypothèse à l’aide d’images statiques ou de vidéos de visages. Or, les récentes études suggèrent que l’emploi de tels stimuli « non conformes » pourraient négliger l’influence du contexte social. La présente étude a cherché à savoir si la tendance à fixer la région des yeux sur le visage persisterait dans un contexte permettant une interaction sociale. Dans l’expérience 1, nous montrons un scénario dans lequel un duo pouvait s’engager dans une interaction sociale, au cours de laquelle leurs mouvements oculaires étaient enregistrés. Dans le cas présent, nous montrons que l’humain a tendance à fixer la région des yeux d’un interlocuteur qui est physiquement présent. Aussi, nous déclarons que que le temps passé par un interlocuteur d’un duo à fixer les yeux de l’autre est un bon indicateur de la durée pendant laquelle le deuxième fixera les yeux. Dans l’expérience 2, nous cherchions à savoir si les participants s’ajustent au niveau de contact visuel initié par un interlocuteur en demandant à un complice de jouer le rôle d’un interlocuteur. Le complice a reçu comme instructions de soit, fixer les yeux de l’observateur ou de balayer le visage entier. Le comportement de regard du complice n’a pas affecté le comportement de regard des observateurs. Nous en concluons que l’humain a effectivement tendance à fixer les yeux lorsque des interlocuteurs peuvent s’engager dans une interaction sociale. De plus, la durée de temps passée à fixer les yeux est duo-dépendante, mais pas facilement manipulée lorsqu’on dicte le comportement de regard d’un des deux interlocuteurs.