Introduction

A number of different eye-tracking systems are currently available to researchers. Despite the apparent similarity of the physical components of some of the current systems, important differences in their technical specifications mean that some systems are more suited to certain types of research questions than others. Some eye-tracking systems are presented as easy and effortless to use, and do indeed allow users to run an experiment with almost no knowledge of the system and limited understanding of experimental design. However, eye-trackers make advanced physiological measurements, and therefore knowing a system, its technical aspects and their impact on the quality of the data collected is crucial for determining whether a study will be feasible with a particular system (Holmqvist et al., 2011). Thus, choosing the right equipment and understanding its features and capabilities are of paramount importance for conducting high-quality eye-tracking research. The aim of this chapter is to provide an overview of the different types of eye-tracking systems available and of the main technical specifications that need to be considered when planning eye-tracking research. For practical reasons, it is not possible to review all eye-tracking systems currently available, so we will review the features of some of the most commonly used ones: EyeLink II, 1000/1000 Plus, Portable Duo (SR Research, www.sr-research.com); Tobii T/X series, Tobii Pro TX300 (Tobii, www.tobii.com); and iView X RED, RED and X Hi-Speed series (SensoMotoric Instruments (SMI), www.smivision.com).

Types of Eye-Trackers

Performing an eye-tracking experiment involves acquiring a continuous record of the eye's position relative to a visual stimulus, such that the recorded eye gaze is matched with the time course of the stimulus (Sedivy, 2010). For example, if we present participants with two images – x and y – we may want to know at specific time points where a participant was looking: at image x, image y or somewhere else. Most of the commercially available eye-trackers currently on the market are video-based systems, with eye-movements monitored by an infra-red camera directed at a participant's eyes. Recordings of the locations of the eyes are based on pupil-only detection or pupil and corneal reflection detections, with the latter being the predominant method. Pupil with corneal reflection tracking has the advantage of compensating for small head movements, which has made it the dominant tracking method (Holmqvist et al., 2011).

Designing an Experiment

Designing and building an eye-tracking study can be a daunting task. In this chapter we will walk through the practicalities of planning, creating and running a study from beginning to end. The specifics of how to actually build an experiment vary according to the type of study we want to run and the eye-tracking system we are using. Our aim is not to provide a ‘how-to’ guide as such, but to talk about the key components and general issues that will be important to most studies. We will also give a brief introduction to the three systems that were discussed in Chapter 2: SR Research EyeLink (www.sr-research.com), Tobii (www.tobii.com) and SMI (www.smivision.com) eye-trackers. As with most things, the best way to learn how to build an eye-tracking study on any of these systems is to dive in and try for ourselves. A good approach is to first look at an existing, similar study – if possible observing an experienced user as he/she builds an experiment – to see the ‘dos’ and ‘don'ts’, then to try editing and building a study of our own.

The first question we should ask is whether eye-tracking is even the best methodology to use. It is tempting to adopt new technologies even when a different approach might produce more useful data, so we should first consider the alternatives to decide whether this is the best option for our research purposes. As with any experiment in any field, the research question should inform the methodology and design, rather than the other way around. Reading extensively around our topic will show us how other researchers have chosen to study it, and this should help us to decide whether eye-tracking is a good choice. Assuming that this decision has been made (and that the answer was yes), this chapter addresses some of the things we will need to consider to get started. We will begin by describing the process of preparing stimuli and building an experiment, then discuss factors such as selecting suitable participants, setting up and calibrating the eye-tracker, and collecting and saving data.

First of all, we need to try to conceptualise our experiment as clearly as we can. Normally we start with a set of general questions that are not specific to eye-tracking, but which are important when designing any language study.

Introduction

It's clear that reading is a visual task, so it is fairly obvious why eye-tracking is an appropriate methodology to study it. But how (and why) might we use eye-tracking to study listening?

When listening, most of us tend to look at something. In experimental contexts, we can manipulate what participants see so that the visual input has various relationships to what is heard and then explore how these relationships influence looking patterns. In many authentic contexts the auditory and the visual input also have a close connection to each other. In audio storybooks for children, the pictures correspond to the story that children hear. In other situations, like watching a film, the correspondence between the auditory and the visual input might seem less clear – what characters talk about doesn't necessarily relate to the visual environment. Nevertheless, eye-tracking can be informative in a number of ways. For instance, if listeners generally look at someone who is speaking, what happens when subtitles are present on the screen? Eye-tracking can address this question, as well as many others in the context of listening and multimodal input, which will be the focus of this chapter.

The discussion of reading in Chapter 4 showed that there are limits on the eyes’ visual acuity. Vision is sharp around the fovea but decreases moving outwards. The eyes move (saccade) three to four times a second, with brief pauses (fixations), to bring new regions into an area of good visual acuity. The constraints on the visual system apply to viewing static and moving images, just as they did to reading, which means that the basic pattern of saccades and fixations is observed across a wide range of tasks. Importantly, eye-movements in both reading and viewing are thought to be under cognitive control (Luke and Henderson, 2016; Rayner, 2009; but see Vitu, 2003, for an alternative viewpoint). This means that the cognitive processes related to perception, memory and language influence the ‘where and when’ of eye-movements (Luke and Henderson, 2016). For applied linguists, what is important is that the eyes move quickly in response to the linguistic input and the perceptual features of what is being looked at.

What Is Eye-Tracking?

Have you ever watched someone's eyes as they read or look at a scene? What do you notice about the eyes’ behaviour from simple observation? There are descriptions of eye-movements, as well as what they might tell us, that go back as far as Aristotle (Wade, 2010). However, this kind of observation only provides a limited understanding of eye-movements, and historically led to the erroneous conclusion that eyes ‘glide over scenes and to alight on objects of interest, which they would fix with unmoved accuracy’ (Wade and Tatler, 2011, p. 37). It was not until the late nineteenth and early twentieth centuries, when early eye-trackers were developed, that we gained insight into the rapid discontinuous nature of eye-movements.

The ‘twitches’ and ‘jerks’ of the eye – in other words their movements – are what we commonly refer to as saccades. The interval between the eyes’ movements, when the eyes ‘stop’, are called fixations. Both are a type of automatic, physiological response, which means that they are not under our conscious control (Rayner et al., 2012). In reading, saccades do not always move the eye forward in a text. About 10–15 per cent of the time, readers move their eyes back (regress) to previously encountered sections of text. These backward movements are referred to as regressions. Regressions can be short or long. Short regressions are usually due to overshooting a target.