A study on visual preference for social stimuli in typical Ecuadorian preschoolers as a contribution to the identification of autism risk factors

The goal of this study was to examine the visual preference towards socially salient stimuli, using a low-cost eye-tracking device in a group of typically developing (TD) Ecuadorian preschoolers aged 11 to 60 months, from rural and urban areas, and from families with low to high socioeconomic status (SES). Series of original stimuli inspired by those used in Western experiments on the early detection of Autism Spectrum Disorder (ASD) were proposed in two eye-tracking tasks. Two types of movements (human vs. object) were presented in task 1, and dynamic speaking faces in task 2. Parental perceptions of the adaptability of the low-cost eye-tracking device used here were also investigated through a questionnaire. The analyses of mean fixation times showed a visual preference for human movements compared to moving objects whatever age, residency location or SES. In task 2, visual preference for the mouth’s area compared to the eyes’ area was observed in specific conditions, modulated by residency location and SES but not by age. The analyses of the parental perception indicated that the eye-tracking technique is well accepted. The findings suggest that these stimuli, along with the experimental procedure and low-cost eye-tracking device used in the present study may be a relevant tool that can be used in clinical settings as a contribution to the early identification of at-risk factors of ASD in low- and middle-income contexts.

www.nature.com/scientificreports/ and East Asian adults. These eye-tracking findings are coherent with the cultural expectations that direct eye contact is considered rude in many Asian cultures. Studies on visual orientation towards social stimuli are not reported in Latin American countries. Henceforth, there is a need to investigate the contribution of a low-cost eye-tracking display to identify at-risk factors of ASD in low-and middle-income settings, such as rural clinical practices and community services. Sociocultural factors, together with other variables, such as unfamiliarity with test situations and with screen-based tools, particularly in the case of eye-tracking, could also influence the visual fixation results and the successful implementation of assessment techniques in culturally diverse settings.
Objectives and hypotheses. The main objective was to examine the visual preferences towards socially salient stimuli using a low-cost eye-tracking device in a group of typically developing Ecuadorian preschoolers from rural or urban areas and from families with low to high socioeconomic status (SES). Series of original stimuli inspired by those used in Western experiments on the early detection of ASD were proposed to examine visual preferences in two tasks. Two types of movements (human vs. object) were presented in task 1 and inner dynamic talking faces were presented in task 2. Parental perceptions of the adaptability of the eye-tracking setting were investigated through a short questionnaire.
We hypothesized that a visual preference for human movements compared to object movements in task 1 and for eyes compared to mouth should be observed. We also investigated whether these visual preferences are modulated by age, residency location or SES factors. Finally, we hypothesized that the eye-tracking setting would be well accepted and adapted to study child development in a non-academic context in a low-and middle-income contexts.

Participants.
A final sample of 37 typically developing children (aged from 15.2 to 59 months; M = 33.37, SD = 15.22) reached the inclusion criteria for eye-tracking tasks. Participants were recruited through public announcements in the town of Quito and its rural surroundings. The absence of developmental disorders in these children was controlled with a questionnaire addressed to parents. Children younger than 31 months were assessed by the Spanish version of M-chat, validated for a Mexican population 41 . Children older than 32 months were assessed with the SCQs translated to Spanish 42 . Both instruments were self-administered by the primary caregivers, assisted by the researcher when requested. The Bayley Scales of Development was used to assess the level of communication in children younger than 42 months. The VABS 43 was used to assess the scale of language on children older than 42 months. Information on hearing and visual status, considered to be a factor of exclusion in cases of impairment or deficit in this study, was collected through a parental questionnaire.
Parents were also interviewed using a socio-demographic questionnaire. These socio-demographic characteristics are described in Table 1.
Parents gave their informed written consent for their child to take part in the present study, which was conducted in accordance with the Declaration of Helsinki and whose experimental protocol had been approved by the Ethics Committee of the University of Geneva. Parents of acting children and adults playing the role of a "caregiver, " provided an informed written consent for participation and for publication on an online openaccess publication.
Experimental procedure. Children accompanied by a parent were hosted either in a separate room provided by the city's playschools or in a communal house in rural areas. In both cases, the experiment took place in a closed room where the light could be controlled and adapted to the optimal conditions for the eye-tracking setting. Each child was tested individually. Children were placed on an adjustable height seat with a headrest behind their neck to maintain head stability. Parents stood behind the children and were required not to move for the duration of the experiment in order to avoid distracting or moving the child or having their eyes captured by the eye-tracker. Children were seated about 60 cm from a device (Dell screen of 47.5 × 30 cm with a spatial resolution of 1680 × 1050 pixels) equipped with an "EyeTribe" (https:// theey etribe. com/ theey etribe. com/ about/ index. html) eye-tracker (an adjustable table was used to set the height of the stimulation screen so that the participant's eyes were positioned in the center of the screen when looking straight ahead as recommended). Eye-tracking settings, such as this one, provide non-invasive measurement of fixation times using corneal reflection to detect oculomotor movements. The system records near-infrared reflections of both eyes at 60 Hz (accuracy = 0.5°, spatial resolution = 0.1°). Accuracy and precision of outcomes for image-based studies comparable to those of well-established eye-tracking devices 44,45 , and it is considered to be a valuable resource for research in cognitive psychology 12,46 . The eye-tracker was positioned at approximately 45 cm from the child's face, ensuring that its top side was aligned with the underside of the stimulation screen 45 .
The experiment consisted of four blocks: a first calibration phase followed by two visual preference tasks (randomly ordered) and by a short parental questionnaire to investigate parents' perception of the experiment. A calibration with 9 points was required to verify that the eye motion could be directed to different areas. Quality of calibration was displayed on the screen. If a child's calibration was poor, the procedure was repeated as necessary. Participants were required to look to the circular target that was displayed at different locations of the screen on a blank background, for approximately 2 s each. This phase was repeated until a satisfactory calibration was achieved: it was only once the eye-tracking device was successfully calibrated that the two visual preference tasks (lasting approx. 35 s and about 1 min each) could begin. Both tasks were designed to test the tendency to look longer (in terms of mean fixation times) at socially salient stimuli. The order of presentation of these two tasks was balanced to limit a possible drop in attention (order effect). At the beginning of each task, a children's character with a music-box sound was presented for 4 s in order to bring the children's attention to the screen. www.nature.com/scientificreports/ The same character and sound had already been presented for 4 s at the very beginning of the experiment (before the calibration phase) for the same purpose. Both types of stimuli were recorded with the participation of South American actors and were previously tested in a population of 15 adults. These stimuli were judged as neutral to an Ecuadorian population. Before leaving, the parents were quickly questioned about their perception of the experimental procedure that their child had just gone through, and the child received a wooden-puzzle (worth approx. USD 3) at the end of the experiment as a thank you for participating.
Measures. Task 1: visual preference for human vs. object movements. In the first task, we used a procedure inspired by the stimuli used by Pierce et al. 21 . In this study, authors projected, side by side, a series of pairs of images on a screen to children aged between 12 and 42 months with and without ASD. To study the visual preference for social versus non-social scenes, we used a paradigm of preferential fixation time, quantified by the mean fixation time on socially prominent scenes (faces of children in movement) compared to repetitive movements of the object. Although the preference for human movements cannot be fundamentally questioned when using a stimulus based on irregular movements across the screen, such as geometric motion used by Pierce et al. 21 , we assumed that the pattern of movements used to study the preference could affect the outcomes. For this reason, we created a stimulus using spinning objects that are centered. In this case the AOIs were designed to fill approximately the same proportion of space in the screen for human and for non-human (object) movements. Hence, only the head region of actors was shown to participants. Round objects were chosen for object movements aiming to make the comparison more precise. Social scenes were recorded in a neutral environment with white background and consisted of videos of children moving their heads, faces, and mouths naturally. Object movement scenes consisted of the repetitive movement of a bicycle wheel or a fan (cf. Figure 1 for screenshots or see videos 1 to 6 in online associated Supplementary material). The final stimulus was made up of six pairs Table 1. Parents socio-demographics characteristics. a The family's socioeconomic status (SES) was approximated using the price of the preschool as a proxy ranging from one to two (the highest SES).  Figure 1. Serie of original stimuli used to quantify the visual preferences for two types of movements (human vs. object). Each pair of stimuli was presented for 5 s., for a total task duration of 34.5 s. www.nature.com/scientificreports/ of scenes distributed horizontally across the screen, each one measuring approximately 9 × 10 cm, that corresponded to the size of the AOIs, with a distance of 3 cm between the 2 stimuli and was presented on a dark grey background. The visual angles of the stimuli were Horizontal: 20°, Vertical: 10°. No sound accompanied the stimuli. The duration of the presentation of each pair of scenes was of 5 s. The presentation of human and object stimuli was counterbalanced between the left and right sides of the screen in order to control a spatial preference. The presentation of social stimuli was also alternated between male and female actors. The influence of low-level elements, such as the color of faces was not controlled, to maintain their ecological character. Likewise, the speed of movement of children's faces has been kept to preserve the natural character of these movements and may not correspond to the speed of movement of objects that were filmed and presented at their usual speed. As far as possible, the color temperatures on both sides of the screen were standardized. Between each sequence, we presented a turning wheel in the middle of a black background screen (approx. 2 × 2 cm) for 1.5 s., and after 3 sequences a character with a music-box sound was presented, aiming to bring participants' attention back to its center. The total duration of the presentation of the sequences was of 34.5 s, with 25 frames per second.
Task 2: visual preference for the facial region of the mouth vs. the eyes. In the second task, we followed a procedure inspired by the studies of Jones, Carr and Klin 34 and Jones and Klin 35 . The stimuli we designed consisted of five video-clips (available in online associated Supplementary material), each presenting a naturalistic scene of an actress addressing directly to the camera playing the role of a "caregiver" and singing nursery rhymes in Spanish, intended to attract the child's attention (Fig. 2). The actress was filmed in an ecological setting simulating a room with paintings in the background. The final stimulus was made up of 5 scenes distributed vertically across the screen, each one measuring approximately 16 × 13 cm and was presented on a dark grey background. The sizes of the 2 AOIs was 6.5 cm. × 3.5 cm, with a distance between AOIs of 4.5 on average depending on the size of the face. The visual angles of the stimuli were Horizontal: 15°,Vertical: 12°. The videos were projected for a total duration of just over 1 min, with 25 scenes per second. The sound was output through speakers not visible to the children. The audio duration was on average 11.5 s. The volume of auditory stimuli presented did not exceed 60 dBA.
Parental perception of the adaptability of eye-tracking to assess young children with a short questionnaire. In order to study the parent's perception of the eye-tracking as an assessment technique, parents were asked (1) if they felt comfortable during the experiment with their child, (2) if they would recommend a friend to participate in the assessment with their child 11 , and (3) if they had questions related to the experiment. The response options for the first two questions were "yes", "no" and "maybe".
Data processing. All oculometric data (fixation times in milliseconds) were extracted using Open Gaze and Mouse Analyzer (Ogama-version 5.0), developed by researchers at the Free University of Berlin. In each of the two eye-tracking tasks, only data from participants who looked at the stimulation screen for more than 50% of the time were analyzed. The first task was completed by 43 children (data from three participants were excluded due to procedural failures) but only 30 of them looked at the screen long enough during the task (more than 50% of the time) to include their data in our analyses. This sample (N = 30; 15 females) consisted of full-term born (at least 37 weeks of gestation) children aged 15.2 to 59.2 months (M = 32.7, SD = 10.3) divided into the following three age groups: 11-24 months (n = 8); 25-36 months (n = 11); 37-60 months (n = 11). Their family socioeconomic status (SES) was calculated as a proxy of the cost paid monthly by the preschool and classified in "low to medium-low (LML) SES" (n = 9) and "medium-high to high (HMH) SES" (n = 21). They were living in urban and rural areas (see Table 1 for more details).
Although our 46 participants completed the second task, only the data from the 27 children who spent more than 50% of the task duration looking at the stimulation screen were analyzed. The descriptive characteristics of this sample are as follows: children (N = 27; 16 females) aged 15.2 to 48.5 months (M = 32.4, SD = 9.7) were all born at least at 37 weeks of gestation. These children were distributed among three age groups as follows: 11-24 months (n = 6); 25-36 months (n = 12); 37-60 months (n = 9). They came from families with LML (n = 7) and HMH (n = 20) socio-economic status and were living in urban (n = 17) and rural (n = 10) areas (cf. Table 1).

Statistical analysis.
Statistical analyses were computed using TIBCO Statistica 13.2 computer software. We used an alpha level of 0.05 for all statistical tests. www.nature.com/scientificreports/ Firstly, we implemented a non-factorial plan with two AOIs (human vs. object movements in task 1 and mouth vs. eyes' areas in task 2) as independent variable. In the first task, AOIs corresponded to the size of each stimuli (i.e. 9 × 10 cm). In the second task, we defined two AOIs corresponding to the eye's region and the mouth region, each measuring 6.5 × 3.5 cm, with a distance between the two AOIs of 4.5 cm on average (depending on the size of the face).
We aimed to analyze the mean fixation time, assuming that longer fixation duration is a measure of a more pronounced visual engagement. Focusing, at first, on the analysis of the mean fixation time of all the subjects (in average) towards the two AOIs, we conducted a T-test for independent samples. We then focused on the analysis of the mean fixation time within-subjects towards the two AOIs by performing a T-test for dependent samples in order to explore whether each participant looked at both AOIs equally.
Secondly, we conducted a series of factorial analyses to test the influence of the following three factors on the visual preference for socially salient stimuli: age group (11-24, 25-26 vs. 27-60 months), residency location (rural vs. urban areas), and SES (LML vs. HMH). For this purpose, three separate analyses were conducted: (1) a T-test for dependent samples was used to test whether the individuals in each group category thus formed looked significantly longer at one of the two stimuli presented simultaneously during the two tasks (intra-group difference), (2) inter-group comparisons were performed (using either a T-test for independent samples or an One-way ANOVA according to the number of factor modalities) to test whether individuals in the different group categories had equivalent or significantly different fixation times towards each stimulus, and (3) mixed factorial ANOVAs finally enabled us to test the interaction between the stimuli type/area (in task 2) and the group categories.

Results
Visual preference in task 1 (human vs. object movements). The first result for the whole sample indicated that the average of the mean fixation time on human movements was significantly higher (M = 1683.67, SD = 390.11), compared to object movements (M = 774.85, SD = 394.65). Indeed, the analysis considering the mean fixation time of all the subjects toward each of the two AOIs independently was significant [t(58) = 8.97, p < 0.001] as well as the analysis of the mean fixation time within-subjects towards the two AOIs, t(29) = 6.987, p < 0.001.
We found similar results in our three factorial analyses: children in all three age groups also looked at human movements longer than at object movements (see intra-group difference in analysis b in Table 2), as do children living in rural and urban areas (see intra-group difference in analysis c in Table 2) and those from families with low to medium-low and medium-high to high SES (see intra-group difference in analysis d in Table 2).
Moreover, the different modalities of these factors did not seem to influence the visual preferences of the individuals they clustered: the results showed that there were no significant differences in the fixation times for each stimulus between each group category (see inter-group difference analyses in Table 2).  For the results of the factorial analyses, we first noted that there was no visual preference for the mouth compared to the eyes' regions in any of the three age groups (see intra-group difference in analysis b in Table 3). Children from the three different age groups (i.e. 11-24 months [t(5) = 1.967, p = 0.106], 25-36 months [t(11) = 0.472, p = 0.646] and 37-60 months [t(8) = 0.939, p = 0.375]) all looked at both AOIs equally: neither the differences in fixation times on each AOI between the three age groups (see inter-group difference in analysis b in Table 3) Finally, neither inter-group differences nor interactions tested were significant for the last two factors (see analyses c and d in Table 3).
Taken together, these results indicate that the AOIs in faces have an impact on the visual preference of the whole sample, suggesting a preference for the mouth region. However, this preference is not significant when we study the visual preferences within participants. In task 1, the results suggest that the type of movement has a significant impact on visual preferences, as well as when this preference is studied within-subjects. Specifically, our results suggest that when Ecuadorian preschoolers are presented with dynamic images of faces, there is a significant preference for the mouth region, but only in the whole group, an effect that does not appear when the fixation time between the two regions is compared within-subjects. On the contrary, when participants have the choice, they significantly prefer human movements against object movements, and this preference is maintained in the whole sample as well as within-subjects. Finally, in both tasks the developmental factor does not seem to have an impact, as no difference appears between age groups.   www.nature.com/scientificreports/ Parental perceptions. We used the parent's perceptions as a measure of the feasibility of eye-tracking studies in Ecuador. 144 responses were analyzed. All caregivers reported a feeling of comfort during the experiment. Almost all participants (99%) reported they would agree to invite a friend or family member to participate in a similar study with their child. The question relative to caregivers' questions or comments about the experiment was analyzed through Content Analysis, a method used to identify emerging categories from participants' answers 47,48 . The responses and comments to this open-ended question were transcribed verbatim and classified in at least one of the three emerging categories. The first category (10% of participants) was related to the way an eye-tracker functions (e.g., "How does it know where my child looks?", "How does it work?"). The second category (26% of participants) was related to the stimuli or about the calibration procedure, (e.g., "Why is there a child next to a wheel, "What do the points mean?"). The third category referred to an absence of questions (65% of participants). Interestingly, a homogenous sub-group of participants, from a rural area of residency and from families with low to middle-low SES (n = 17) did not ask any questions nor provide any comments.

Discussion
The overarching aim of this study was to define the adaptability of the eye-tracking technique to semi-experimental and non-experimental settings in low-and middle-income contexts. To address this question, we used stimuli and procedures that have been previously used in occidental studies in order to quantify preferential visual attention towards socially salient stimuli. We also studied parents' perceptions of the eye-tracking application. The analyses of mean fixation times in task 1 showed a visual preference for human movements compared to object movements, whatever age, residency location or SES. This preference is maintained as well within-subjects confirming previous findings that used similar stimuli 21,22,49 .
The analyses of mean fixation times in task 2 showed that the AOIs on faces inner features have a significant impact on visual preference for the mouth's region in the whole sample, but not when visual preference is studied within-subjects. These outcomes may be interpreted as a part of the mixed results reported by the literature. A first explanation to the variability of results related to the preference for the mouth's region is the influence of the stimuli type and design used 13,50 .
An alternative explanation corresponds to complex developmental milestones. While some results indicate that two-year-olds with ASD look less at the eyes' area and more at the mouth's area compared to TD children 34 , others studies have found typical levels of looking times towards the eyes and reduced looking times towards the mouth in TD children aged 2 to 4 36 . In another study comparing ASD to TD preschoolers, Nakano et al. 51 found that the majority of TD participants spontaneously directed their gaze towards the mouth of a person speaking. Based on this result, the authors suggested that this could be a necessary step in typical development. A finding that agrees with outcomes indicating that greater amounts of fixation time on the mother's mouth during live interactions at the age of 6 months predicted higher levels of expressive language at the age of 18 months 52 . Alternative explanations to the preference for the mouth area, such as the size of the mouth and teeth of actors cannot be ruled out.
The analyses of the parental perception showed that eye-tracking setting is very well accepted and adapted. All the participants felt comfortable with the procedure during the experiment. Although 99% reported they would agree to invite a friend or family member to participate with their child, the rate of adherence explained by the presence of the researcher could not be ruled out. Parents' comments and interests were mostly related to the calibration procedure (dots moving around on the screen) and about the meaning of the stimuli. A small number of questions were related to the technology or the way eye-tracking works to capture the direction of the gaze. Interestingly an important proportion of participants did not provide comments or questions when given the opportunity. Most of the participants in this case belonged to the LML SES, suggesting that the presence of the experimenter could have had an influence.
The first limit of this study relates to the stimuli used in task 2. Some scenes may have shown more salient in certain features of the face, depending on the actors, that are difficult to control (for example the size of the mouth relative to the size of the eyes, size of the teeth, etc.) A high rate of calibration difficulties has been observed in this sample, suggesting the need of using a character who can more easily attract children's attention instead of calibration points. Visual acuity, which is a potential confounder, has also been difficult to control. The results of this study must be viewed with caution due to the small size of the sample. The current study was limited to one low-and middle-income setting; it is unlikely, however, that other understated elements not detected in this study would create major barriers for the implementation of eye-tracking procedures in a context such as Ecuador.
In spite of these limitations, this work provides some evidence on the feasibility of eye-tracking studies in typically developing young children in a low-and middle-income setting. It also suggests that a paradigm of visual preferences for human and non-human movements would be valid for use among the Ecuadorian population, in urban and rural areas, and in different socioeconomic contexts.
The study of social orientation in typical development, is necessary in order to provide a sound basis for explorations of atypical developments, such as those characterizing children with ASD. However, the use of shared protocols and experimental paradigms allowing data collection across different cultures and socio-economic settings would be necessary to determine whether the utility of eye-tracking technique extends beyond the mapping of socio-cognitive development, and could inform on its potential clinical utility in the identification of early markers in different socio-cultural contexts.

Data availability
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.