The view from the road: the contribution of on-road glance-monitoring technologies to understanding driver behavior.

Using glance-monitoring technologies for on-road studies is an excellent way to investigate driver behaviors in an ecologically valid setting. Recent advances in glance-monitoring technologies have made it possible to conduct on-road studies of drivers' glance behavior that heretofore were simply not possible. Yet it is not always easy to determine which glance-monitoring technology to use for a particular application. Here, we first identify the generic capabilities of the various glance-monitoring technologies. We then describe how particular glance-monitoring technologies have been used in the field to (a) identify the skill deficiencies of novice and older drivers, (b) evaluate the effectiveness of training programs that are designed to reduce deficits in these skills, and (c) address interface issues both inside (e.g., collision warning systems) and outside (e.g., yield markings) the vehicle. The limitations and advantages of on-road eye-tracking and the associated glance-monitoring technologies are identified throughout. A comparison, where possible, is made between the results of on-road eye-tracking studies of drivers' behaviors and the results of those studies conducted in the laboratory. Overall, the use of appropriate on-road glance-monitoring technologies has greatly enhanced our theoretical understanding of why drivers behave the way they do, and this knowledge has paved the way for significant improvements in road user safety.

The effects of focused attention training on the duration of novice drivers' glances inside the vehicle.

Several studies have documented that the failure of drivers to attend to the forward roadway for a period lasting longer than 2-3 s is a major cause of highway crashes. Moreover, several studies have demonstrated that novice drivers are more likely to glance away from the roadway than the experienced drivers for extended periods when attempting to do a task inside the vehicle. The present study examines the efficacy of a PC-based training programme (FOrward Concentration and Attention Learning, FOCAL) designed to teach novice drivers not to glance away forthese extended periods of time. A FOCAL-trained group was compared with a placebo-trained group in an on-road test, and the FOCAL-trained group made significantly fewer glances away from the roadway that were more than 2 s than the placebo-trained group. Other measures indicated an advantage for the FOCAL-trained group as well. Statement of relevance: Distracted driving is increasingly a problem, as cell phones, navigation systems, and other in-vehicle devices are introduced into the cabin of the automobile. A training programme is described that has beentested on the open road and can reduce the behaviours that lead to crashes caused by the distracted driving.

Can younger drivers be trained to scan for information that will reduce their risk in roadway traffic scenarios that are hard to identify as hazardous?

Younger drivers (18-21 years) are over-involved in crashes. Research suggests that one of the reasons for this over-involvement is their failure to scan areas of the roadway for information about potential risks in situations that are hazardous, but not obviously so. The primary objective of the present study is to develop and evaluate a training program that addresses this failure. It was hypothesised that PC-based hazard anticipation training would increase the likelihood that younger drivers would scan for potential hazards on the open road. In order to test this hypothesis, 12 trained and 12 untrained drivers' eye movements were measured as they drove a vehicle on local residential, feeder and arterial roads. Overall, the trained drivers were significantly more likely to gaze at areas of the roadway that contained information relevant to the reduction of risks (64.4%) than were the untrained drivers (37.4%). Significant training effects were observed even in situations on the road that were quite different from those shown in training. These findings have clear implications for the type of training of teen drivers that is necessary in order to increase their anticipation of hazards.

Can novice drivers be trained to scan for information that will reduce their likelihood of a crash?

Sixteen year old drivers are involved in 10.3 fatal crashes per 100 million vehicle miles, a rate almost double that of 18 year olds and almost eight times that of 45-64 year olds, who are the safest group of drivers. Crash rates are particularly higher during the first month of licensure and decline rapidly for about six months and 1000 miles and then much more slowly for at least two years, consistent with a typical learning curve. Research indicates that drivers who have their learner's permit or are just newly licensed have particular difficulties identifying areas of a scenario from which hidden risks could emerge. Standard driver education programs do not appear to address these difficulties adequately. This suggests that some alternative form of driver training could reduce the crashes, either in the classroom or on the road. A PC based program designed to teach drivers to recognize risks early on is shown to improve their awareness of hazards, both on an advanced driving simulator and on the road.

Semantic codes are not used in integrating information across eye fixations in reading: evidence from fluent Spanish-English bilinguals.

The question of whether meaning can be extracted from unidentified parafoveal words was examined using fluent Spanish-English bilinguals. In Experiment 1, subjects fixated on a central cross, and a preview word was presented to the right of fixation in parafoveal vision. During the saccade to the parafoveal preview word, the preview was replaced by the target word, which the subject was required to name. In Experiment 2, subjects read sentences containing the target word, and, as in the naming task, a preview word was replaced by the target word when the subject's saccade crossed a boundary location. In both experiments, preview words were identical to the target word, translations, orthographic controls for the translations, or unrelated words in the opposite language. In both experiments, the preview benefit from the translation conditions was no greater than would be predicted by the orthographic similarity of the preview to the target. Hence, the data indicated that subjects obtained no useful semantic information from words seen parafoveally that enabled them to identify them more quickly on the subsequent fixation.

Eye movement control in reading: word predictability has little influence on initial landing positions in words.

We examined the initial landing position of the eyes in target words that were either predictable or unpredictable from the preceding sentence context. Although readers skipped over predictable words more than unpredictable words and spent less time on predictable words when they did fixate on them, there was no difference in the launch site of the saccade to the target word. Moreover, there was only a very small difference in the initial landing position on the target word as a function of predictability when the target words were fixated which is most parsimoniously explained by positing that a few programmed skips of the target word fell short of their intended target. These results suggest that low-level processing is primarily responsible for landing position effects in reading.

The role of phonological codes in integrating information across saccadic eye movements in Chinese character identification.

Prior research has generally assumed either that phonological codes do not contribute to Chinese character identification or that they do so only through a look-up process at the character level. In 3 experiments, a homophone seen parafoveally aided the identification of a target character that was fixated following an eye movement to the preview location. Moreover, high-frequency phonetically regular characters were named faster than high-frequency, phonetically irregular characters. Thus, both lexical and sublexical phonological codes of Chinese characters are involved early in the process of character identification. Orthographic information from the preview was also used in character identification, as orthographically similar previews facilitated target identification as well. The evidence for the extraction of semantic information from parafoveal previews was mixed, as synonym previews facilitated in Experiment 2 but not in Experiment 1.

The role of morphological constituents in reading Finnish compound words.

The processing of transparent Finnish compound words was investigated in 2 experiments in which eye movements were recorded while sentences were read silently. The frequency of the second constituent had a large influence (95 ms) on gaze duration on the target words, but its influence was relatively late in processing: A clear effect only occurred on the probability of a third fixation. The frequency of the whole compound word had a similar influence on gaze duration (82 ms) and influenced eye movements at least as rapidly as did the frequency of the second constituent. These results, together with an earlier finding that the frequency of the first constituent affected the first fixation duration, indicate that the identification of these compound words involves parallel processing of both morphological constituents and whole-word representations.

Activation of phonological codes during eye fixations in reading.

Two experiments addressed the issue of whether phonological codes are activated early in a fixation during reading using the fast-priming technique (S. C. Sereno & K. Rayner, 1992). Participants read sentences and, at the beginning of the initial fixation in a target location, a priming letter string was displayed, followed by the target word. Phonological priming was assessed by the difference in the gaze duration on the target word between when the prime was a homophone and when it was a control word equated with the homophone on orthographic similarity to the target. Both experiments demonstrated homophonic priming with prime durations of about 35 ms, but only for high-frequency word primes, indicating that lexicality was guiding the speed of the extraction of phonological codes early in a fixation. Evidence was also obtained for orthographic priming, and the data suggest that orthographic and phonological priming effects interact in a mutually facilitating manner.

The effects of "neighborhood size" in reading and lexical decision.

The effects of neighborhood size ("N")--the number of words differing from a target word by exactly 1 letter (i.e., "neighbors")--on word identification was assessed in 3 experiments. In Experiments 1 and 2, the frequency of the highest frequency neighbor was equated, and N had opposite effects in lexical decision and reading. In Experiment 1, a larger N facilitated lexical decision judgments, whereas in Experiment 2, a larger N had an inhibitory effect on reading sentences that contained the words of Experiment 1. Moreover, a significant inhibitory effect in Experiment 2 that was due to a larger N appeared on gaze duration on the target word, and there was no hint of facilitation on the measures of reading that tap the earliest processing of a word. In Experiment 3, the number of higher frequency neighbors was equated for the high-N and low-N words, and a larger N caused target words to be skipped significantly more and produced inhibitory effects later in reading, some of which were plausibly due to misidentification of the target word when skipped. Regression analyses indicated that, in reading, increasing the number of higher frequency neighbors had a clear inhibitory effect on word identification and that increasing the number of lower frequency neighbors may have a weak facilitative effect on word identification.

Extraction of information to the left of the fixated word in reading.

The present experiment used 2 different eye-contingent display change techniques to determine whether information is extracted from English text even when it is to the left of the currently fixated word. Preview display changes were during the 1st saccade entering the target word region, whereas postview display changes were during the 1st saccade leaving that region. Previews and postviews were either identical, related, or unrelated to the target word. "Wrong" information in the target-word region affected reading even when that information was seen only after readers were fixating to the right of that region: When readers skipped the target word, such information caused readers to regress to the target word more; when readers initially fixated the target word, such information increased "2nd-pass" processing time on the target region. The data suggest that readers often still attend to a word after it is skipped and that when readers fixate a word, they occasionally attend to the word after they have begun to fixate the next word.

The activation of phonology during silent Chinese word reading.

The role of phonology in silent Chinese compound-character reading was studied in 2 experiments using a semantic relatedness judgment task. There was significant interference from a homophone of a "target" word that was semantically related to an initially presented cue word whether the homophone was orthographically similar to the target or not. This interference was only observed for exact homophones (i.e., those that had the same tone, consonant, and vowel). In addition, the effect was not significantly modulated by target or distractor frequency, nor was it restricted to cases of associative priming. Substantial interference was also found from orthographically similar nonhomophones of the targets. Together these data are best accounted for by a model that allows for parallel access of semantics via 2 routes, 1 directly from orthography to semantics and the other from orthography to phonology to semantics.

Eye movement control in reading: accounting for initial fixation locations and refixations within the E-Z Reader model.

Reilly and O'Regan (1998, Vision Research, 38, 303-317) used computer simulations to evaluate how well several different word-targeting strategies could account for results which show that the distributions of fixation locations in reading are systematically related to low-level oculomotor variables, such as saccade distance and launch site [McConkie, Kerr, Reddix & Zola, (1988). Vision Research, 28, 1107-1118]. Their simulation results suggested that fixation locations are primarily determined by word length information, and that the processing of language, such as the identification of words, plays only a minimal role in deciding where to move the eyes. This claim appears to be problematic for our model of eye movement control in reading, E-Z Reader [Rayner, Reichle & Pollatsek (1998). Eye movement control in reading: an overview and model. In G. Underwood, Eye guidance in reading and scene perception (pp. 243-268). Oxford, UK: Elsevier; Reichle, Pollatsek, Fisher & Rayner (1998). Psychological Review, 105, 125-157], because it assumes that lexical access is the engine that drives the eyes forward during reading. However, we show that a newer version of E-Z Reader which still assumes that lexical access is the engine driving eye movements also predicts the locations of fixations and within-word refixations, and therefore provides a viable framework for understanding how both linguistic and oculomotor variables affect eye movements in reading.

The time course of phonological, semantic, and orthographic coding in reading: evidence from the fast-priming technique.

The present experiment employed the fast-priming paradigm in reading (Sereno & Rayner, 1992), in which sentences are silently read while eye-movement-contingent changes are made on a specified target region. In this paradigm, when readers fixate on a specified target word region, a prime word is encountered for a brief duration at the beginning of the fixation and then it is replaced by a target word. Three types of primes were employed: homophones, semantically related, and orthographically similar, and five prime durations were employed: 29, 32, 35, 38, and 41 msec. The primary finding was that significant homophone priming was obtained at prime durations ranging from 29 to 35 msec, whereas significant semantic priming occurred only at the 32-msec prime duration. In contrast, significant orthographic priming occurred at all prime durations. These findings indicate that phonological codes are activated during an eye fixation at least as rapidly as semantic codes. An explanation for the pattern of events is suggested using the framework of an activation-verification model.

Reading Finnish compound words: eye fixations are affected by component morphemes.

The role of morphemic processing in reading was investigated in 2 experiments in which participants read sentences as their eye movements were monitored. The target words were 2-morpheme Finnish compound words. In Experiment 1, the length of the component morphemes was varied and word length was held constant, and in Experiment 2, the uniqueness of the initial morpheme was varied and the rated familiarity and length of the word were held constant. The length of the initial morpheme influenced the location of the second fixation on the target word and the pattern of fixation durations (although it had a negligible influence on the gaze duration of the word). The frequency of the initial morpheme influenced the duration of the first fixation on the target word, had a substantial effect on the gaze duration, and also influenced the location of the first and second fixations on the target word. Subsidiary analyses indicated that these effects were unlikely to stem from orthographic factors such as bigram frequency.

Unspaced text interferes with both word identification and eye movement control.

Subjects read either normal text, text in which the space information between words was absent (either spaces were removed filled with x), or text in which spaces were preserved but the words were flanked by x. In two experiments, reading rate decreased by approx. 50% when space information was not available, suggesting that reading unspaced text is relatively difficult. The removal of space information increased the effect of word frequency on the fixation times for selected target words, indicating that word identification was interfered with by the lack of spaces. In addition, removal of space information influenced the initial landing positions on words, indicating that eye movement control was affected by the absence of spaces. Further analyses were conducted that explored the relationship between these two effects.

The effects of neighborhood frequency in reading and lexical decision.

To test the effect of the frequency of orthographic "neighbors" on the identification of a printed word, two sets of words were constructed (equated on the number of neighbors, word frequency, and number of letters); in one set, the words had no higher frequency neighbors and in the other set, they had at least one higher frequency neighbor. Identification was slower for the latter set. In Experiment 1, this was indexed by longer response times in a lexical decision task. In Experiment 2, the target words were embedded in sentences, and slower identification was indexed by disruptions in reading: more regressions back to the words with higher frequency neighbors and longer fixations on the text immediately following these words. The latter results indicate that a higher frequency neighbor affects relatively late stages of lexical access, an interpretation consistent with both activation-verification and interactive activation models.

Evidence for the use of assembled phonology in accessing the meaning of printed words.

The role of assembled phonology in visual word recognition was investigated using a task in which participants judged whether 2 words (e.g., PILLOW-BEAD) were semantically related. Of primary interest was whether it would be more difficult to respond "no" to "false homophones" (e.g., BEAD) of words (BED) that are semantically related to target words than to orthographic controls (BEND). (BEAD is a false homophone of BED because-EAD can be pronounced /epsilon d/.) In Experiment 1, there was an interference effect in the response time data, but not in the error data. These results were replicated in a 2nd experiment in which a parafoveal preview was provided for the 2nd word of the pair. A 3rd experiment ruled out explanations of the false homophone effect in terms of inconsistency in spelling-to-sound mappings or inadequate spelling knowledge. It is argued that assembled phonological representations activate meaning in visual word recognition.

Phonological codes and eye movements in reading.

A number of recent studies using eye movement data have yielded evidence suggesting that phonological codes are activated early in an eye fixation. However, experiments reported by M. Daneman and E. Reingold (1993; M. Daneman, E. M. Reingold, & M. Davidson, 1995) yielded data that led them to argue that phonological codes are primarily activated after lexical access has occurred. In this study, 3 experiments were carried out that were conceptually similar to those of M. Daneman and E. Reingold, and the resulting data supported the position that phonological codes are activated very early in an eye fixation.

Toward a model of eye movement control in reading.

The authors present several versions of a general model, titled the E-Z Reader model, of eye movement control in reading. The major goal of the modeling is to relate cognitive processing (specifically aspects of lexical access) to eye movements in reading. The earliest and simplest versions of the model (E-Z Readers 1 and 2) merely attempt to explain the total time spent on a word before moving forward (the gaze duration) and the probability of fixating a word; later versions (E-Z Readers 3-5) also attempt to explain the durations of individual fixations on individual words and the number of fixations on individual words. The final version (E-Z Reader 5) appears to be psychologically plausible and gives a good account of many phenomena in reading. It is also a good tool for analyzing eye movement data in reading. Limitations of the model and directions for future research are also discussed.