A Handbook of Process Tracing Methods

2nd Edition

Michael Schulte-Mecklenbeck, Anton Kuehberger, Joseph G. Johnson, Michael Schulte-Mecklenbeck, Anton Kuehberger, Joseph G. Johnson

• Publisher: Routledge
• Year: 2019
• ISBN:
  9781351662758
• Edition: 2
• Topic: Psychologie
• Subtopic: Psychologie cognitive et cognition

A Handbook of Process Tracing Methods demonstrates how to better understand decision outcomes by studying decision processes, through the introduction of a number of exciting techniques. Decades of research have identified numerous idiosyncrasies in human decision behavior, but some of the most recent advances in the scientific study of decision making involve the development of sophisticated methods for understanding decision process—known as process tracing. In this volume, leading experts discuss the application of these methods and focus on the best practices for using some of the more popular techniques, discussing how to incorporate them into formal decision models.

This edition has been expanded and thoroughly updated throughout, and now includes new chapters on mouse tracking, protocol analysis, neurocognitive methods, the measurement of valuation, as well as an overview of important software packages. The volume not only surveys cutting-edge research to illustrate the great variety in process tracing techniques, but also serves as a tutorial for how the novice researcher might implement these methods.

A Handbook of Process Tracing Methods will be an essential read for all students and researchers of decision making.

1

Eye Fixations as a Process Trace

J. Edward Russo

The eye is our preeminent sense for acquiring information from the external environment, and nearly all such acquisition is accomplished when the eye fixates on an object. The eye has evolved so that its movement to the next fixation point is the fastest of all external information acquisition behaviors. Its unique value has led humans to rely on it for nearly all intellective tasks. Without vision can one imagine the development of writing, and without writing, the development of civilization?¹

One such intellective task is decision making, or more generally judgment and decision making (JDM). Almost every decision involves the acquisition of visual information. The main exceptions occur when all the needed information is purely auditory like music or is retrieved from long-term memory, that is, via internal rather than external search. Tracking the eye during a decision yields a series of fixations on different units of the available information.

Such a sequence of fixations brings unique advantages. Because it is the least effortful of information acquisition responses (“metabolically cheap” in the biological phrasing of Spivey, Richardson, & Dale, 2009, p. 237), a fixation sequence provides a detailed trace of the information acquired. Indeed, an eye movement to reacquire information is so cheap that it can sometimes substitute for the use of working memory as a storage and retrieval system (Droll & Hayhoe, 2007; Maxcey-Richard, & Hollingworth, 2013). Further, eye fixations can be recorded nonreactively, and they are difficult to censor. Both properties derive in part because eye movements are fast and largely automatic. No manual response can compare in detail, cost, and veridicality, even pointing a computer mouse (although this information acquisition technology is widespread and useful; see Chapters 6–10).

At the same time, there are barriers. The greatest of these was the availability of an eye-tracking apparatus. That barrier has changed substantially over the last decade, with accurate eye trackers now costing only a few hundred dollars (Lejarraga, Schulte-Mecklenbeck, & Smedema, 2016; Titz, Scholz, & Sedlmeier, 2017). Indeed, the supply of eye-tracking apparatus probably exceeds the demand—for the first time in history. The biggest remaining barrier is the task of interpreting the recorded sequence of fixations. A fixation tells researchers where individuals are looking, not what they are thinking. However, it is the latter that is needed for the development and testing of explanations of observed JDM phenomena.

Fundamentals of Eye Movements and Measurement Technology

Types of Eye Movements

Saccades

For JDM studies and cognition more generally, one type of eye movement is the most relevant: saccades (Henderson, 2006). These are rapid, voluntary movements of widely different amplitude from one object of regard to another. Historically, the broad reliance on a sequence of saccades to accomplish most cognitive tasks was not appreciated until the end of the 19th century. Instead, the eye was thought to move more continuously, especially in such common visual tasks as scanning scenes, search, and reading. Unlike most of our movements, we do not notice our own eyes move. For example, you have no awareness of your own saccades as you read this sentence, but you can clearly observe another reader’s eyes leap from one word to another and down to the next line. Such movement was labeled “par saccade” by the French physiologist Javal, and can be translated roughly as movement “by fits and starts”, with the single word “saccade”² corresponding to something like the English words “jerk” or “leap” (Wade, 2007).

See Wade and Tatler (2005) for comprehensive histories of eye movement research and the much earlier detailed review in Chapter 2 of Yarbus (1967). The performance capabilities of saccades are well documented (e.g., Carpenter, 1988; Becker, 1991). These movements are rapid, reaching a peak velocity of 700° per sec for longer movements. When acceleration and deceleration are incorporated, a 2° saccade that is typical in reading takes about 25 msec, while a 6° movement that might occur in scanning a visual scene requires about 35 msec.

An important characteristic of saccades is the “suppression” of vision during movement (Matin, 1974; Thiele, Henning, Buischik, & Hoffman, 2002).³ During a saccade, essentially no information can be acquired, certainly not the recognition of alphanumeric characters in typical JDM tasks.⁴

Among voluntary movements, saccades are so effortless that they usually go unnoticed. Thus, we recognize errors in other voluntary movements like motor “missteps”, but there are no “misfixations”. We may fixate something anticipated to be interesting and then find that it is not, but the error is one of prediction, not one of incorrectly executing a saccadic movement. Even when the movement is aimed imperfectly, the control system of the eye rapidly corrects its aim and accomplishes the necessary adjustment so effortlessly that the correction is no more recognized than the original saccade. In contrast, a misstep while walking, even when it is corrected in time to avoid tripping, is often a more conscious action, likely eliciting embarrassment when done in a public situation.

Non-Saccadic Movements

There are other movements of the eyes besides saccades. In vergence movements, the eyes jointly focus on a single point or object of regard. They contribute to both visual acuity and depth perception. Smooth pursuit movements allow the eye to track an object that follows a continuous trajectory. Such movements maintain the target at a fixed point on the retina, usually the fovea, in order to attain sufficient resolution for such ordinary actions as walking over uneven terrain. Nystagmus movements enable constant focus on an object of regard when the head moves. Such compensatory movements are essential to vision as observers move through their environment. There are also very small movements (tremors, drifts, and flicks) that will, in most cases, not concern JDM researchers. For a more detailed description of the varieties of eye movements, see Kowler (1995; 2011) or Duchowski (2017, especially Chapter 4).

Methods for Tracking Eye Movements

Since the late 1800s researchers have wanted to track eye movements (Yarbus, 1967; see also Alpern, 1962). Early techniques were often intrusive, for example, making plaster of Paris casts that fit over the eyeball (Delabarre, 1898). Over the next century, the success of eye fixations as process data was largely driven by better instruments and techniques. For instance, direct contact with the eye and other methods typically required immobilization of the head. These methods gave way to ones in which the eye-tracking apparatus can be mounted on the head, today achieved with unobtrusive glasses (see Chapter 2). The concern of JDM researchers is mainly functional, what the available apparatus can do, and not how it does it. Nonetheless, it should not be forgotten that sometimes the “how” constrains the “what”, even in JDM studies. For instance, a glasses-mounted eye tracker used while a consumer traverses a supermarket aisle offers limited ability to identify which elements of package facings are fixated or even exactly which packages are fixated if the latter are small.

Hardware

The most popular eye-tracking system uses a video camera, pointed at the eye, often illuminated with infrared light, to record its corneal reflection. Some devices also record the reflected image off the back surface of the lens for additional resolution. Other methods currently in use are electro-oculography, a reflecting contact lens on the sclera (the white of the eye), and tracking the limbus which is the border between the iris and sclera. Thorough reviews of the available eye-tracking technologies are provided by Duchowski (2017), Holmqvist et al. (2011) and Liversedge, Gilchrist, and Everling (2011).

The ideal system, which various laboratories and commercial firms have essentially achieved, provides excellent resolution and high recording frequency, as well as being head mounted (or unobtrusively placed) so that it is able to track the eye even as the head moves. The latter capability is especially valuable in dynamic environments like those faced by airplane pilots or automobile drivers (e.g., Lee, 2009; Friedrich, Russwinkel, & Mohlenbrink, 2017). There are also some common decisions, such as supermarket purchases, where the consumer’s progress through a rich and changing environment requires allowance for head motion.

Nonetheless, current systems are more than adequate for nearly all JDM studies, often not even requiring researchers to modify their preferred visual display of the stimulus.⁵

Software

Software has become an increasingly important part of the technology of eye movement tracking. It has greatly facilitated the recording of eye position. It has also enabled the automation of both the distinction of fixations from movements (e.g., Blingnaut, 2008) and the identification of areas of interest (e.g., Hessels, Kemner, van den Boomen, & Hooge, 2016). Current software also enables gaze-initiated changes in the stimulus (Duchowski, Cournia, & Murphy, 2004; Glaholt, Rayner, & Reingold, 2012; McConkie & Rayner, 1975; Rayner, 1975; Rosen, 1975). That is, as a function of where a participant is looking, another part of the visual display is changed. For instance, to test the number of letters in the “perceptual span” while reading, a letter N spaces ahead is altered. Only if this is noticed, does its position lie within the perceptual span.⁶ These systems enable various other stimulus manipulations like a moving window in which all but the “window” is occluded or masked (Bertera & Rayner, 2000; Bradshaw, Nettleton, Wilson, & Nathan, 1984; Franco-Watkins & Johnson, 2011; Franco-Watkins, Davis, & Johnson, 2016; Rayner, Smith, Malcolm, & Henderson, 2009). They also permit the serial presentation of stimuli to the fovea (i.e., without eye movements) and the magnification of a portion of the stimulus, termed foveated imaging/rendering (Miellet, O’Donnell, & Sereno, 2009). Some of the most intriguing current and potential JDM research using eye fixations involves the software of eye-tracking systems.

Object of Regard

The most direct use of eye fixations is to identify the objects of regard in a visual environment. To simplify the presentation, this section is partitioned into applications where objects of regard are treated as individual entities and those where a distribution of fixations over multiple objects is the primary focus.

Individual Objects of Regard

A sense of the variety and importance of applications of eye fixations can be conveyed by considering content areas only marginally related to JDM. For example, Mello-Thoms, Nodine, and Kundel (2002; Kundel & Nodine, 1978) found all cases when there was visual evidence of breast cancer (a “lesion” present in an X-ray) but radiologists failed to report it. In 59% of those cases, the lesion had been fixated (yet not reported as seen). Similarly, Fischer, Richards, Berman, and Krugman (1989) showed that when viewing cigarette advertisements, three out of seven adolescents never fixate the mandated warning.

Predicted Fixations

Eye fixation data can be used to test an object of regard that has been predicted from theory or from some standard of comparison. For instance, in an unusual test of the “just world hypothesis”, eye fixations revealed that good (bad) behavior influenced the fixation to the corresponding visual location even before the appearance of the actual good (bad) outcome (Callan, Ferguson, & Bindemann, 2013). The success of many magic tricks depends on the misdirection of visual attention, something naturally verified by recording an observer’s eye fixations. In an unusual application of predicted fixations, Tatler and Kuhn (2007) tested the trick of making a lighted cigarette “disappear”. Briefly, attention is drawn to one hand while the other hand removes the cigarette from the mouth of the magician (who was Kuhn). Then, just before it reaches the table in front of him, the lit cigarette is dropped onto his lap. By the time that the subject’s gaze returns to the hand that held the cigarette, the latter has “disappeared”. Video recording of the subject’s gaze confirmed the misdirection of attention away from the cigarette when it was dropped out of sight. That is, eye fixations away from a crucial object of regard, the dropped cigarette, verified the misdirection hypothesis. In a broader statement of the value of predicted objects of regard, Henderson (2017) argues that fixations in viewing complex scenes are better understood as the result of knowledge-based ...