WEBVTT 00:00:00.000 --> 00:00:06.090 align:middle line:90% 00:00:06.090 --> 00:00:08.730 align:middle line:84% Hi, my name is Natalia Kartushina. 00:00:08.730 --> 00:00:10.770 align:middle line:84% I am associate professor at MulitLing, 00:00:10.770 --> 00:00:14.010 align:middle line:84% which is a centre on multilingualism in society 00:00:14.010 --> 00:00:17.280 align:middle line:84% across the lifespan here at the University of Oslo. 00:00:17.280 --> 00:00:18.450 align:middle line:90% And my name is Audun. 00:00:18.450 --> 00:00:22.200 align:middle line:90% I am a PhD fellow at MulitLing. 00:00:22.200 --> 00:00:25.830 align:middle line:84% So yeah, we've seen that eye tracking pupillometry 00:00:25.830 --> 00:00:30.120 align:middle line:84% can be used to examine so many different phenomena in adults. 00:00:30.120 --> 00:00:33.870 align:middle line:84% But here, Audun, you do research with babies. 00:00:33.870 --> 00:00:38.760 align:middle line:84% So you use eye tracking to understand how babies develop. 00:00:38.760 --> 00:00:42.120 align:middle line:84% Yeah, so I study language acquisition. 00:00:42.120 --> 00:00:44.880 align:middle line:84% And I'm interested in figuring out 00:00:44.880 --> 00:00:48.090 align:middle line:84% what are the mechanisms and factors in the environment that 00:00:48.090 --> 00:00:52.785 align:middle line:84% are important for babies when they learn a language. 00:00:52.785 --> 00:00:54.910 align:middle line:84% So in the end, they will become competent speakers. 00:00:54.910 --> 00:00:57.035 align:middle line:84% But before that, there's a lot of interesting steps 00:00:57.035 --> 00:00:59.730 align:middle line:84% that happen along the way, really early on. 00:00:59.730 --> 00:01:05.560 align:middle line:84% And because infants, before they can talk, 00:01:05.560 --> 00:01:10.270 align:middle line:84% they can't really tell us what they know about the world. 00:01:10.270 --> 00:01:13.410 align:middle line:84% And we can't have them do decision-making tasks 00:01:13.410 --> 00:01:16.720 align:middle line:84% that you can do with adults or older children. 00:01:16.720 --> 00:01:20.310 align:middle line:84% So eye tracking and pupillometry is a very clever way 00:01:20.310 --> 00:01:25.500 align:middle line:84% to sort of figure out what they know about the world. 00:01:25.500 --> 00:01:29.100 align:middle line:84% Yeah, so do you mean by measuring infants look 00:01:29.100 --> 00:01:33.150 align:middle line:84% at the screen, we can learn about how they 00:01:33.150 --> 00:01:34.770 align:middle line:90% develop their language skills? 00:01:34.770 --> 00:01:36.760 align:middle line:90% Yeah, exactly. 00:01:36.760 --> 00:01:39.570 align:middle line:84% Yeah, so for example, with eye tracking, 00:01:39.570 --> 00:01:45.930 align:middle line:84% where we are mostly interested in gaze patterns and fixations, 00:01:45.930 --> 00:01:50.550 align:middle line:84% we can sort of imply where attention 00:01:50.550 --> 00:01:53.280 align:middle line:84% is focused by measuring where they 00:01:53.280 --> 00:01:56.760 align:middle line:84% look at the screen when they are exposed to different pictures 00:01:56.760 --> 00:01:58.860 align:middle line:90% or sounds. 00:01:58.860 --> 00:02:01.950 align:middle line:84% Whereas for pupillometry, where we measure 00:02:01.950 --> 00:02:05.190 align:middle line:84% the diameter of the pupil, we can rather 00:02:05.190 --> 00:02:11.880 align:middle line:84% infer how much, or the degree of attention. 00:02:11.880 --> 00:02:15.270 align:middle line:84% So for example, if we wanted to figure out 00:02:15.270 --> 00:02:19.530 align:middle line:84% what words does an infant understand before they can 00:02:19.530 --> 00:02:23.160 align:middle line:84% actually start to say them, we could 00:02:23.160 --> 00:02:26.550 align:middle line:84% design a pupillometry experiment where 00:02:26.550 --> 00:02:31.680 align:middle line:84% we show different pictures on a screen 00:02:31.680 --> 00:02:37.140 align:middle line:84% and we play different audio together with these pictures. 00:02:37.140 --> 00:02:41.100 align:middle line:84% So that the infant hears the label of something 00:02:41.100 --> 00:02:43.840 align:middle line:90% that it sees on the screen. 00:02:43.840 --> 00:02:46.470 align:middle line:84% So let's say that the infant, he or she, 00:02:46.470 --> 00:02:49.560 align:middle line:90% sees the picture of a dog. 00:02:49.560 --> 00:02:53.950 align:middle line:84% But we can manipulate the audio to, instead of being like, 00:02:53.950 --> 00:02:56.010 align:middle line:84% well, look at the dog, the infant 00:02:56.010 --> 00:02:58.380 align:middle line:90% can hear look at the car. 00:02:58.380 --> 00:03:00.840 align:middle line:84% Oh yeah, just something absolutely different, 00:03:00.840 --> 00:03:03.720 align:middle line:90% no a relationship with the word. 00:03:03.720 --> 00:03:06.730 align:middle line:90% [NORWEGIAN] 00:03:06.730 --> 00:03:07.230 align:middle line:90% 00:03:07.230 --> 00:03:09.150 align:middle line:90% Exactly, doesn't make sense. 00:03:09.150 --> 00:03:12.540 align:middle line:84% So that's what we would call a mismatch trial. 00:03:12.540 --> 00:03:15.420 align:middle line:84% So there's a mismatch between what she hears 00:03:15.420 --> 00:03:17.160 align:middle line:90% and what she sees. 00:03:17.160 --> 00:03:21.690 align:middle line:84% And you can then measure the pupil dilation as 00:03:21.690 --> 00:03:25.620 align:middle line:84% compared to a trial where the child would see a dog 00:03:25.620 --> 00:03:29.310 align:middle line:84% and also hear it being named a dog. 00:03:29.310 --> 00:03:32.760 align:middle line:84% And then the difference then in the diameter 00:03:32.760 --> 00:03:34.950 align:middle line:84% of the pupil between these different trials 00:03:34.950 --> 00:03:39.990 align:middle line:84% can then help us see that, well, if the infant or child knows 00:03:39.990 --> 00:03:43.800 align:middle line:84% the word "dog," she should really 00:03:43.800 --> 00:03:46.140 align:middle line:84% have some kind of mismatch response 00:03:46.140 --> 00:03:48.600 align:middle line:84% when something is not right between the picture 00:03:48.600 --> 00:03:50.640 align:middle line:90% and the audio. 00:03:50.640 --> 00:03:55.890 align:middle line:84% Wow, that sounds very elegant and kind of very easy. 00:03:55.890 --> 00:04:00.270 align:middle line:84% But isn't it challenging to work with babies? 00:04:00.270 --> 00:04:01.260 align:middle line:90% They move. 00:04:01.260 --> 00:04:03.120 align:middle line:84% They might not want to do what you 00:04:03.120 --> 00:04:06.090 align:middle line:84% ask them to do as you would do with the adults. 00:04:06.090 --> 00:04:07.480 align:middle line:90% Their behaviour. 00:04:07.480 --> 00:04:09.000 align:middle line:90% How do you work with them? 00:04:09.000 --> 00:04:10.718 align:middle line:84% They behave like babies typically do. 00:04:10.718 --> 00:04:12.510 align:middle line:84% Also when they come to visit us in the lab, 00:04:12.510 --> 00:04:13.385 align:middle line:90% they're still babies. 00:04:13.385 --> 00:04:17.310 align:middle line:84% They're not like magically paying attention or doing what 00:04:17.310 --> 00:04:18.700 align:middle line:90% whatever we want them to do. 00:04:18.700 --> 00:04:22.500 align:middle line:84% So you have to be a bit clever in terms of your design 00:04:22.500 --> 00:04:26.080 align:middle line:84% and set up, especially important with pupillometry. 00:04:26.080 --> 00:04:29.280 align:middle line:84% So for example, one thing that is really important 00:04:29.280 --> 00:04:33.810 align:middle line:84% is that you have a secure seating. 00:04:33.810 --> 00:04:37.260 align:middle line:84% So when they sit in front the eye tracker, with an adult, 00:04:37.260 --> 00:04:39.510 align:middle line:90% you typically use a chin rest. 00:04:39.510 --> 00:04:42.840 align:middle line:84% So that you will remain very still with the head. 00:04:42.840 --> 00:04:45.008 align:middle line:84% And it's easier to get reliable data. 00:04:45.008 --> 00:04:46.800 align:middle line:84% But with infants, you can't really put them 00:04:46.800 --> 00:04:48.420 align:middle line:90% into a chin rest, of course. 00:04:48.420 --> 00:04:53.220 align:middle line:84% So instead, we have this kind of different type of eye tracker 00:04:53.220 --> 00:04:55.200 align:middle line:90% that is-- 00:04:55.200 --> 00:04:57.420 align:middle line:84% allows for free movement of the head. 00:04:57.420 --> 00:05:00.720 align:middle line:84% So that the infant can actually be seated in the parent's lap 00:05:00.720 --> 00:05:04.290 align:middle line:84% or in a car seat if that's necessary. 00:05:04.290 --> 00:05:05.800 align:middle line:90% And they can still move around. 00:05:05.800 --> 00:05:08.650 align:middle line:84% But if we take a sticker and then we 00:05:08.650 --> 00:05:10.880 align:middle line:84% place it on the infant's forehead, 00:05:10.880 --> 00:05:13.640 align:middle line:84% yeah, typically we ask the parent to-- 00:05:13.640 --> 00:05:14.380 align:middle line:90% A target. 00:05:14.380 --> 00:05:16.930 align:middle line:90% Yeah, yeah. 00:05:16.930 --> 00:05:18.160 align:middle line:90% It's not always too easy. 00:05:18.160 --> 00:05:22.360 align:middle line:84% But in the end, you can get the sticker to be put there. 00:05:22.360 --> 00:05:24.730 align:middle line:84% And then that will measure the distance 00:05:24.730 --> 00:05:28.250 align:middle line:84% to the eye tracker relative to the head movement. 00:05:28.250 --> 00:05:29.380 align:middle line:90% So they can still be free. 00:05:29.380 --> 00:05:32.210 align:middle line:84% And you will still get good data. 00:05:32.210 --> 00:05:34.150 align:middle line:84% So that's one thing that you would really do. 00:05:34.150 --> 00:05:35.830 align:middle line:84% And you do that when you calibrate, 00:05:35.830 --> 00:05:37.550 align:middle line:90% before you start the experiment. 00:05:37.550 --> 00:05:39.310 align:middle line:84% So you can do both like a calibration 00:05:39.310 --> 00:05:42.140 align:middle line:84% and then a validation of the calibration. 00:05:42.140 --> 00:05:45.230 align:middle line:84% So we make sure that everything is correct. 00:05:45.230 --> 00:05:49.090 align:middle line:84% So by having the infant look at different directions, 00:05:49.090 --> 00:05:51.880 align:middle line:84% different areas of the screen, different corners. 00:05:51.880 --> 00:05:53.530 align:middle line:84% And then you start to experiment. 00:05:53.530 --> 00:05:56.350 align:middle line:84% Yeah, that sounds-- wow, there are so 00:05:56.350 --> 00:05:58.960 align:middle line:84% many steps to think about before you can run 00:05:58.960 --> 00:06:00.430 align:middle line:90% an experiment with the babies. 00:06:00.430 --> 00:06:04.360 align:middle line:84% And I guess given very limited attention span in babies, 00:06:04.360 --> 00:06:06.490 align:middle line:84% the experiment must be very short, right? 00:06:06.490 --> 00:06:08.890 align:middle line:84% Because you can't spend a lot of trials. 00:06:08.890 --> 00:06:11.560 align:middle line:84% And even with a short experiment, 00:06:11.560 --> 00:06:14.920 align:middle line:90% a child can get bored. 00:06:14.920 --> 00:06:17.200 align:middle line:84% So what do you do if the child is not 00:06:17.200 --> 00:06:19.450 align:middle line:84% looking at the screen and he just doing-- 00:06:19.450 --> 00:06:21.160 align:middle line:84% or she is doing just something else? 00:06:21.160 --> 00:06:22.540 align:middle line:90% What do you do in those cases? 00:06:22.540 --> 00:06:23.642 align:middle line:90% That happens a lot. 00:06:23.642 --> 00:06:24.850 align:middle line:90% That happens a lot, for sure. 00:06:24.850 --> 00:06:27.460 align:middle line:84% Because I mean even if you were in a room with very 00:06:27.460 --> 00:06:31.660 align:middle line:84% little other sort of interest areas, 00:06:31.660 --> 00:06:35.200 align:middle line:84% an infant can still be interested in his or her socks 00:06:35.200 --> 00:06:38.410 align:middle line:84% or the experimenter in the back, or the parent, whatever. 00:06:38.410 --> 00:06:40.130 align:middle line:90% So that happens. 00:06:40.130 --> 00:06:43.450 align:middle line:84% So what we typically do to try to fixate their attention back 00:06:43.450 --> 00:06:48.850 align:middle line:84% towards the stimuli at the screen is that we programme 00:06:48.850 --> 00:06:51.280 align:middle line:90% in these attention getters. 00:06:51.280 --> 00:06:51.790 align:middle line:90% Oh, yeah. 00:06:51.790 --> 00:06:55.870 align:middle line:84% Yeah, so if the infant looks away from the screen, 00:06:55.870 --> 00:07:00.820 align:middle line:84% the experimental script will automatically play like 00:07:00.820 --> 00:07:05.740 align:middle line:84% a bull's eye kind of animation with colourful wheels, 00:07:05.740 --> 00:07:09.130 align:middle line:84% something like that that's very attractive, 00:07:09.130 --> 00:07:13.870 align:middle line:84% combined with maybe chirping sounds or chirping-- 00:07:13.870 --> 00:07:15.760 align:middle line:90% Both visual and audio. 00:07:15.760 --> 00:07:18.160 align:middle line:84% Yeah, and actually, this works really well 00:07:18.160 --> 00:07:20.320 align:middle line:84% to have them fixate back at the screen. 00:07:20.320 --> 00:07:21.610 align:middle line:90% And then they hear this. 00:07:21.610 --> 00:07:24.250 align:middle line:84% And then you continue with the experiment. 00:07:24.250 --> 00:07:27.820 align:middle line:84% And even after all these precautions, 00:07:27.820 --> 00:07:31.300 align:middle line:84% typically, you would still have pretty noisy data 00:07:31.300 --> 00:07:33.670 align:middle line:90% from an infant. 00:07:33.670 --> 00:07:37.840 align:middle line:84% So then how do we go about analysing such data? 00:07:37.840 --> 00:07:39.190 align:middle line:90% Yeah, you're right. 00:07:39.190 --> 00:07:43.870 align:middle line:84% I think one thing that we could do and that we should do 00:07:43.870 --> 00:07:47.470 align:middle line:84% is to define all these inclusion and exclusion 00:07:47.470 --> 00:07:48.550 align:middle line:90% criteria beforehand. 00:07:48.550 --> 00:07:50.440 align:middle line:84% Before we start analysing the data, 00:07:50.440 --> 00:07:53.740 align:middle line:84% we can decide what is a valid trial, 00:07:53.740 --> 00:07:56.770 align:middle line:84% how much time the child needs to attend to the screen 00:07:56.770 --> 00:07:59.672 align:middle line:84% so the child can be considered as a valid trial. 00:07:59.672 --> 00:08:00.880 align:middle line:90% To be included in the final-- 00:08:00.880 --> 00:08:02.560 align:middle line:90% [INTERPOSING VOICES] 00:08:02.560 --> 00:08:04.190 align:middle line:90% And also what's the good child? 00:08:04.190 --> 00:08:08.740 align:middle line:84% So how many trials do we need to consider 00:08:08.740 --> 00:08:10.960 align:middle line:84% how many trials the child needs to contribute 00:08:10.960 --> 00:08:13.120 align:middle line:84% with for the child to be included 00:08:13.120 --> 00:08:15.280 align:middle line:84% in the final sample for the analysis. 00:08:15.280 --> 00:08:19.120 align:middle line:84% And all these things can be preregistered beforehand. 00:08:19.120 --> 00:08:21.700 align:middle line:84% So we have less degrees of freedom 00:08:21.700 --> 00:08:24.760 align:middle line:84% later, once we have to process the data. 00:08:24.760 --> 00:08:29.230 align:middle line:84% And so then we can use all these exciting designs 00:08:29.230 --> 00:08:33.309 align:middle line:84% to examine infants' language development. 00:08:33.309 --> 00:08:37.000 align:middle line:84% And of course all the domains, social cognition, 00:08:37.000 --> 00:08:39.280 align:middle line:84% physical cognition, emotional cognition, 00:08:39.280 --> 00:08:45.460 align:middle line:84% all these really important, cool steps before infant becomes-- 00:08:45.460 --> 00:08:46.120 align:middle line:90% An adult. 00:08:46.120 --> 00:08:47.740 align:middle line:90% Yeah, yeah, yeah. 00:08:47.740 --> 00:08:50.190 align:middle line:90% Yeah, absolutely. 00:08:50.190 --> 00:09:00.000 align:middle line:90%