Participants
A total of 33 university students (22 women; age M = 22.2, S.D. = 3.29) participated in the study. These participants were all right-handed and had normal or lens-corrected vision, no diagnosis of psychiatric or neurological disorders, and no school attendance issues. The number of participants was calculated considering an alpha of 5%, beta of 90%, and effect size of 0.26 (large). The calculation was made on G*Power® 3.1.9.2 (Buchner, Erdfelder, Faul, & Lang, 2017). The participation of all subjects was voluntary and approved by the university’s Research Ethics Committee. Subjects gave written informed consent and received course credit in return at the end of the procedure.
Adult Dyslexia Checklist
The Adult Dyslexia Checklist (ADC; Vinegrad, 1994) is a questionnaire of 20 items, all of which are related to symptoms of different areas of dyslexia. The items comprise questions in a “yes” or “no” answer format (e.g., “Is map reading or finding your way to a strange place confusing?”). For each item marked in the affirmative, a point is added to the test result.
Although the instrument may indicate the possibility of dyslexia, it is not a diagnostic tool. In other words, the data collected in this test are not sufficient to definitively identify dyslexia. However, the test results have a high indicative value for dyslexia. It would be useful to suggest that subjects with high scores undergo an evaluation with a complete multidisciplinary team (Vinegrad, 1994).
Lexical decision task
The lexical decision task was adapted from Oliveira (2014). We incorporated the feasibility criteria for the application and recording of behavioral responses and ocular movements. Three categories of linguistic items were defined, yielding a total of 216 items: 72 regular words, 36 pseudo-words, and 108 quasi-words. The syllabic structure of the stimuli was counterbalanced among CVCVCV (e.g., Pirata [Pirate]), VCVCV (e.g., Urina [Urine]), CCVCVCV (e.g., Granada [Granada]), and VCCVCV (e.g., Osmose [Osmosis]) structures. The number of letters in the stimuli ranged between 5 and 7 letters, so length had no influence on the processing of the items.
All words used have a medium or high frequency of use in Portuguese, according to the NILC Corpus of the University of São Carlos (http://www.linguateca.pt/ACDC/). We selected words with regular structures and rules. Quasi-words comprised three subtypes of pseudo-words (e.g., Seabra, Dias, Mecca, & Macedo, 2017): quasi-words with visual exchanges, quasi-words with phonological exchanges, and quasi-words with pseudo-homophones. The criteria for the classification of these quasi-word subtypes have been supported in the literature on cognitive models of reading, since errors in the reading of irregular words indicate difficulties in, or the absence of lexical processing (Ellis & Young, 1988). Our categorization is based on that used by Proverbio and Adorni (2008).
Pseudo-words were constructed of sequences of decodable letters and syllables but not derived from real words. For this reason, the frequency values of the bigrams of the task stimuli with 5 and 6 letters were measured according to Justi and Justi (2009).
The task stimuli were created as Joint Photographics Experts Group (JPEG) files with a resolution of 1280 × 720 pixels. The font used was 22-point Calibri in black on a white background. Between each word presented, a fixation point was shown for 2 s (see Fig. 1). The order of the words was randomized.
The participants were instructed to judge whether the word was real and to press the letter “Q” on the keyboard with the left hand if so or “P” with the right hand if not. In front of these letters were marks indicating what the keys meant. Participants were instructed to respond as quickly as possible. Only the behavioral data were used in this research.
Semantic decision task
The semantic decision task was structured to evaluate participants’ ability to judge the ambiguity of written sentences. The task comprised 80 sentences, of which 40 were ambiguous phrases (AMB) and 40 were direct phrases (i.e., unambiguous phrases). Of the direct phrases, 20 were unambiguous sentences with actions related to the subject (ARS) and 20 were unambiguous sentences with actions related to the object (ARO). The sentences had two parts: a first sentence, which gave the context (e.g., “The principal accused the student”), and a second sentence containing the ambiguity or the relation to the subject/object (e.g., “He was processed/He was fired/He was suspended”). The sentences were structured to be the same size with the same number of words (e.g., “The spider attacked the snake. It was poisonous/The spider attacked the snake. It had legs”).
The task stimuli were created in JPEG files with a resolution of 1280 × 720 pixels. The font used was 22-point Calibri on a white background. The stimuli were presented with intervals of 2 s between the participant’s decision and the display of the next sentence. During this interval, a fixation point was presented at the center of the screen (see Fig. 2). The order of the sentences was randomized.
The instructions given to the participants were similar to those for the lexical decision task. The participants were to judge whether the phrase was ambiguous and to press “Q” on the keyboard with the left hand if so or “P” with his right hand if not. In front of these letters were marks indicating what the keys meant. Participants were instructed to react as quickly as possible.
Apparatus
The ocular measurement equipment used was the SensoMotoric Instruments (SMI) RED500 (2014). This equipment, which was connected to a 22″ monitor, allowed the measurement of eye movements. Some of the measures that could be obtained with this equipment were the number of fixations, the total fixation time, the number of saccades, the total time in the trial, and qualitative analyses of ocular patterns, among several others.
The device came with experiment development software, SMI Experiment Center ™, and eye movement analysis software, SMI BeGaze™. It was also compatible with third-party software such as E-Prime, which we used to perform the two experiments. Data collection was performed at 500 Hz. The criteria for identifying fixation and saccades were defined as the default in the SMI BeGaze™ version 3.7.104.
Procedure
The participants came to the laboratory, and the consent terms were explained before they decided whether or not to participate in the research. If they accepted the terms, they completed the ADC and were taken to the room with the eye-tracking equipment. They sat approximately 70 cm from the monitor, which was adjusted to accommodate their physical characteristics. After the participants were positioned, we calibrated the equipment, and the participants then began their first task. The order of the tasks was randomized. For both tasks, participants were given the instructions and started the test when they felt ready. When they had made their judgments about the words or sentences, they pressed the appropriate key on the keyboard in front of them on the monitor table. Between each stimulus presentation, a fixation point was presented at the center of the screen for 2 s. After the test ended, participants received course credit.
Measures
We assessed several variables in this study. These included the percentage of correct items, the average trial time (in microseconds), and the inverse efficiency score (IES), which is the trial time divided by the correct percentage. This latter variable allows the equalization of the time and correct item percentage. Low scores indicate higher efficiency, and higher scores indicate lower efficiency (Bruyer & Brysbaert, 2011). Other variables were the average number of fixations on trial, the average time per fixation (in microseconds), and the percentage of regressive saccades.
Data analysis
The data obtained were submitted to statistical tests that assumed a normal sample distribution. Parametric tests were used because the violation of the normality assumption for samples over 30 is considered unproblematic (Elliott & Woodward, 2007; Ghasemi & Zahediasl, 2012; Pallant, 2001). Cronbach’s alpha was used to analyze the internal consistency of the tasks. In addition, Fleiss’ kappa (Landis & Koch, 1977; Zapf, Castell, Morawietz, & Karch, 2016) was used to assess the inter-rater reliability of the semantic decision task and to confirm the validity of the task. The kappa was calculated with six coders. The coders have experience in the area of neuropsychological assessment and were instructed on the definitions of ambiguous sentences, sentence with ARS, or sentences with ARO before evaluating the semantic decision task. Repeated measures ANOVAs were used to compare the three categories of words and sentences (regular words, pseudo-words, and quasi-words; ambiguous, subject action-related, or object action-related sentences) and their positions in sentences (subject, object, or second sentence). Effect sizes were reported in partial eta-squared, and we calculated their magnitude according to the multiple regression magnitudes (i.e., small < .03, medium < .14, large < .27; Cohen, 1988; Cohen, Cohen, West, & Aiken, 2003; Field, 2009; Watson, 2017). Additionally, stepwise linear regressions were used to identify the factors relevant to semantic decision task efficiency, correct percentage, and average time.