Repetition priming is a phenomenon whereby people respond more quickly to a word when it is preceded by itself, compared to when it is preceded by an unrelated word. Repetition priming has historically been studied using means-based analysis of reaction times where participants make manual responses. In this experiment, we study response times to repeated and new words in an ocular-response task rather than a manual-response task, leading to shorter response times that are closer to what is found in normal reading (Hoedemaker & Gordon, 2014), and we perform ex-Gaussian analysis which allows for localization of effects within the positively-skewed distributions (Balota & Yap, 2011). Although lexical repetitions do not frequently occur within sentences, word recognition studies such as these can inform models of sentence processing.
Thirty-eight native English-speaking adults participated in a contingent-display eye-tracking task while their eye movements were recorded. Trials consist of five evenly spaced words, masked by octothorpes (##) until revealed when the participants’ gaze triggers the invisible boundaries. As the eyes move to the right towards a mask covering the next word, an invisible boundary is triggered which causes the next word to appear on the screen as the eyes reach its location. This technique prevents parafoveal preview and rereading. Because response times in ocular-response tasks are closer to the time spent reading a word (as opposed to the time a participant takes to read and press a button), we are able to obtain a more direct measure of the time spent processing the words without the influence of other processes that may take place as a person executes a button response in a task (Hoedemaker & Gordon, 2014).
Participants were instructed to indicate (via button press at the end of each trial) whether they saw the name of an animal on that trial, and feedback was provided (“Correct” or “Incorrect”). All stimuli were low-frequency English words of various types (nouns, verbs, etc.) ranging from 4-6 letters in length, with log subtitle frequencies ranging from 0.48-1.23. Targets and control primes were matched for length and frequency (to the tenth of a log frequency value). There were 5 practice trials, 60 trials containing an animal name (which were not analyzed), and 160 critical trials (in which the target appeared in position 2, 3 or 4). In half of the critical trials, the target was preceded by itself; in the other half, the target was preceded by an unrelated control prime. There were two counterbalanced lists, such that half of the subjects saw the same target word in each of the two conditions.
We obtained ex-Gaussian parameter estimates for every participant’s target word gaze durations separately in the primed and unprimed conditions using the QMPE v2.18 program for quantile maximum likelihood estimation (Cousineau, Brown, & Heathcote, 2004). Inferential statistics were run on the ex-Gaussian parameter estimates for each participant. There was a significant effect of repetition priming in mu (t(37) = 6.20, p < .001) and in sigma (t(37) = 2.40, p = .021), but not in tau (t(37) = -0.73, p = .470). Results indicate that repetition priming is an automatic head-start process which causes a significantly faster response to repeated words as compared to new words, with a relatively consistent effect size across the two reaction time distributions. Results from manual-response lexical decision tasks have provided some evidence that semantic priming is also a head-start process (Balota, Yap, Cortese, & Watson, 2008); ocular-response tasks have not provided evidence of an effect on mu, instead finding an effect on tau which is consistent with a compound-cue mechanism that only begins when the target is seen (Hoedemaker & Gordon, 2014; 2017).
