Steven G. Luke

Evaluating Significance in Linear Mixed-Effects Models in R

Mixed-effects models are being used ever more frequently in the analysis of experimental data. However, in the lme4 package in R the standards for evaluating significance of fixed effects in these models (i.e., obtaining p-values) are somewhat vague. There are good reasons for this, but as researchers who are using these models are required in many cases to report p-values, some method for evaluating the significance of the model output is needed. This paper reports the results of simulations showing that the two most common methods for evaluating significance, using likelihood ratio tests and applying the z distribution to the Wald t values from the model output (t-as-z), are somewhat anti-conservative, especially for smaller sample sizes. Other methods for evaluating significance, including parametric bootstrapping and the Kenward-Roger and Satterthwaite approximations for degrees of freedom, were also evaluated. The results of these simulations suggest that Type 1 error rates are closest to .05 when models are fitted using REML and p-values are derived using the Kenward-Roger or Satterthwaite approximations, as these approximations both produced acceptable Type 1 error rates even for smaller samples.

Predicting cognitive state from eye movements.

In human vision, acuity and color sensitivity are greatest at the center of fixation and fall off rapidly as visual eccentricity increases. Humans exploit the high resolution of central vision by actively moving their eyes three to four times each second. Here we demonstrate that it is possible to classify the task that a person is engaged in from their eye movements using multivariate pattern classification. The results have important theoretical implications for computational and neural models of eye movement control. They also have important practical implications for using passively recorded eye movements to infer the cognitive state of a viewer, information that can be used as input for intelligent human-computer interfaces and related applications.

Co-registration of eye movements and event-related potentials in connected-text paragraph reading.

Eyetracking during reading has provided a critical source of on-line behavioral data informing basic theory in language processing. Similarly, event-related potentials (ERPs) have provided an important on-line measure of the neural correlates of language processing. Recently there has been strong interest in co-registering eyetracking and ERPs from simultaneous recording to capitalize on the strengths of both techniques, but a challenge has been devising approaches for controlling artifacts produced by eye movements in the EEG waveform. In this paper we describe our approach to correcting for eye movements in EEG and demonstrate its applicability to reading. The method is based on independent components analysis, and uses three criteria for identifying components tied to saccades: (1) component loadings on the surface of the head are consistent with eye movements; (2) source analysis localizes component activity to the eyes, and (3) the temporal activation of the component occurred at the time of the eye movement and differed for right and left eye movements. We demonstrate this methods applicability to reading by comparing ERPs time-locked to fixation onset in two reading conditions. In the text-reading condition, participants read paragraphs of text. In the pseudo-reading control condition, participants moved their eyes through spatially similar pseudo-text that preserved word locations, word shapes, and paragraph spatial structure, but eliminated meaning. The corrected EEG, time-locked to fixation onsets, showed effects of reading condition in early ERP components. The results indicate that co-registration of eyetracking and EEG in connected-text paragraph reading is possible, and has the potential to become an important tool for investigating the cognitive and neural bases of on-line language processing in reading.

Eye movement control in scene viewing and reading: evidence from the stimulus onset delay paradigm.

The present study used the stimulus onset delay paradigm to investigate eye movement control in reading and in scene viewing in a within-participants design. Short onset delays (0, 25, 50, 200, and 350 ms) were chosen to simulate the type of natural processing difficulty encountered in reading and scene viewing. Fixation duration increased linearly with delay duration, and the effect was equivalent for both tasks. Although fixations were longer in scene viewing, the effects of onset delay were highly consistent across tasks. These results suggest that reading and scene viewing share a common mechanism for saccade planning and control.

Context Strengthens Initial Misinterpretations of Text

Three self-paced reading experiments examined the effect of context on interpreting subsequent sentences and in the difficulty of revising initial misinterpretations of subsequent temporarily ambiguous sentences. Target sentences containing noun phrase/sentence (NP/S) coordination ambiguities were preceded by contexts that either did or did not support the preferred, incorrect “NP and NP” interpretation. Online reading times and offline comprehension question responses were the dependent variables. Results suggest that when propositional content of incoming text is consistent with propositional content of the context, readers often hang on to the resulting coherent interpretation even when subsequent input contradicts it. Results also suggest that (a) context affects reading times and final interpretation; (b) when context and comprehension questions bias readers toward the incorrect interpretation, even unambiguous sentences are regularly misinterpreted; and (c) both semantic content and syntactic form of context influence how the context and subsequent text are integrated in memory.

Limits on lexical prediction during reading.

Efficient language processing may involve generating expectations about upcoming input. To investigate the extent to which prediction might facilitate reading, a large-scale survey provided cloze scores for all 2689 words in 55 different text passages. Highly predictable words were quite rare (5% of content words), and most words had a more-expected competitor. An eye-tracking study showed sensitivity to cloze probability but no mis-prediction cost. Instead, the presence of a more-expected competitor was found to be facilitative in several measures. Further, semantic and morphosyntactic information was highly predictable even when word identity was not, and this information facilitated reading above and beyond the predictability of the full word form. The results are consistent with graded prediction but inconsistent with full lexical prediction. Implications for theories of prediction in language comprehension are discussed.

Stable individual differences in saccadic eye movements during reading, pseudoreading, scene viewing, and scene search.

Mean fixation duration and mean saccade amplitude during active viewing tasks differ from person to person. Previous studies have shown that these individual differences tend to be stable across at least some tasks, suggesting that they may reflect underlying traits associated with individuals. However, whether these individual differences are also stable over time has not been established. The present study established stable individual differences in mean fixation duration and mean saccade amplitude across 4 viewing tasks, showed that the observed individual differences are stable over several days, and extended these results to standard deviations of fixation duration and saccade amplitude. The results have implications for theories of eye movement control and for using eye movement characteristics as individual difference measures.

Eye Movement Control during Scene Viewing: Immediate Effects of Scene Luminance on Fixation Durations.

Recent research on eye movements during scene viewing has primarily focused on where the eyes fixate. But eye fixations also differ in their durations. Here we investigated whether fixation durations in scene viewing are under the direct and immediate control of the current visual input. Subjects freely viewed photographs of scenes in preparation for a later memory test while their eye movements were recorded. Using a novel scene degradation paradigm based on a saccade-contingent display change method, scenes were reduced in luminance during saccades ending in critical fixations. Results from two experiments showed that the durations of the critical fixations were immediately affected by scene luminance, with a monotonic relationship between luminance reduction and fixation duration. The results are the first to demonstrate that fixation durations in scene viewing are immediately influenced by the ease of processing of the image currently in view. These results are consistent with the CRISP (a timer-Controlled Random-Walk with Inhibition for Saccade Planning) computational model of saccade generation in scenes, proposing that difficulty in moment-by-moment visual and cognitive processing of the scene modulates fixation durations.

Neural correlates of fixation duration in natural reading: Evidence from fixation-related fMRI.

A key assumption of current theories of natural reading is that fixation duration reflects underlying attentional, language, and cognitive processes associated with text comprehension. The neurocognitive correlates of this relationship are currently unknown. To investigate this relationship, we compared neural activation associated with fixation duration in passage reading and a pseudo-reading control condition. The results showed that fixation duration was associated with activation in oculomotor and language areas during text reading. Fixation duration during pseudo-reading, on the other hand, showed greater involvement of frontal control regions, suggesting flexibility and task dependency of the eye movement network. Consistent with current models, these results provide support for the hypothesis that fixation duration in reading reflects attentional engagement and language processing. The results also demonstrate that fixation-related fMRI provides a method for investigating the neurocognitive bases of natural reading.

Oculomotor inhibition of return in normal and mindless reading

Oculomotor inhibition of return (O-IOR) is an increase in saccade latency prior to an eye movement to a recently fixated location, as compared with other locations. To investigate O-IOR in reading, subjects participated in two conditions while their eye movements were recorded: normal reading and mindless reading with words replaced by geometric shapes. We investigated the manifestation of O-IOR in reading and whether it is related to extracting meaning from the text or is an oculomotor phenomenon. The results indicated that fixation durations prior to a saccade returning to the immediately preceding fixated word were longer than those to other words, consistent with O-IOR. Furthermore, fixation durations were longest prior to a saccade that returned the eyes to the specific character position in the word that had previously been fixated and dropped off as the distance between the previously fixated character and landing position increased. This result is consistent with the hypothesis that O-IOR is relatively precise in its application during reading and drops off as a gradient. Both of these results were found for text reading and for mindless reading, suggesting that they are consequences of oculomotor control, and not of language processing. Finally, although these temporal IOR effects were robust, no spatial consequences of IOR were observed: Previously fixated words and characters were as likely to be refixated as new words and characters.