In the current study we investigated the relationship between Oculomotor Inhibition and WM load. We collected eye tracking data from 26 participants who performed a working memory task. They had to match a probed item to one to three preceding small colored shaped items, see Fig. 1 for an illustration of the task. We extracted epochs triggered by the stimulus onset in a range of -0.3 sec to 3 sec relative to this trigger. First, we calculated the microsaccade rate modulation averaged across participants (N = 26), and the baseline corrected to the mean of the 300 ms pre-stimulus period (see the Methods). Figure 2a shows a typical microsaccade inhibition in response to the first stimulus (time = 0) during the encoding period; the rate first decreased below baseline (MS inhibition), and then increased above baseline at around 300 ms after the stimulus onset, indicating a release from the inhibition. During the WM maintenance period, the rate stabilized slightly above baseline, showing a profound decrease in spontaneous microsaccade activity in the high WM load condition (purple line). Then, from 1.1 sec onward, a gradual reduction in the MS rate occurred, reflecting the recruitment of attention in preparing for the probe stimulus 51. Finally, a second inhibition and release in response to the probe stimulus (time = 1.5 sec), during the retrieval period, showed a stronger and perhaps longer inhibition in the high load condition. To quantify these effects, we calculated the MS rate within separate time windows relative to the stimulus onset, 0-0.5, 0.5–1.1, and 1.5–2.1 sec, to account for the encoding, maintenance, and retrieval WM stages, respectively. The MS rate results, averaged across observers, show a negative trend, depending on the number of items to memorize, in all three memory stages, reflecting stronger OMI with increasing load. Significance was assessed using the Linear Mixed Model (LMM, see the Methods), yielding p = 0.014 in the Encoding stage, p = 0.0026 in the Maintenance stage, and p = 0.0007 in the Retrieval stage, Figs. 2b, 2c, and 2d, respectively. The MS rate differences did not stem from the blink rates, since the latter did not vary significantly between conditions.
In Fig. 3 we plotted Individual data and compared the two extreme conditions of one and three items. The results showed that most observers had a higher MS rate in the one-item condition (red circles), compared with the three-item condition (blue circles) in all three WM stages. Paired t-test assessments yielded significance, p = 0.025 in the Encoding stage, p = 0.003 in the Maintenance stage, and p = 0.001 in the Retrieval stage, Figs. 3a, 3b, and 3c, respectively.
The task accuracy results, averaged across observers, were above 84% for all the conditions, but were far from a ceiling effect, demonstrating that the task was difficult enough to drive differential microsaccade inhibition. The task accuracy was significantly higher in the one-item condition, p = 0.0001, LMM (Fig. 4a). We also wanted to determine whether microsaccade execution during different memory stages may affect the task accuracy. Figure 4b shows that the MS rate in all three memory stages, i.e., encoding, maintenance and retrieval, was higher for the incorrect responses compared with the correct ones, p = 0.07 (n.s.), p = 0.013, and p = 0.018, Paired t-test, with a calculated Cohen’s d effect size, ES = 0.76, ES = 1.07, and ES = 1.02, respectively, along with one excluded observer who made no mistakes. These results indicate that better WM task performance is associated with a lower MS rate, especially during the maintenance phase.
Finally, we plotted the participants’ response times during the retrieval period, measured from the probe onset for microsaccades, blinks, and psychophysics, shown in Fig. 4c. We computed the microsaccade reaction time (msRT, see the Methods) as well as the blink RT, as the first microsaccade/blink released from inhibition in a time window ranging from 1.5 to 3 seconds. The ms/blink RT and the participants’ RT show a positive trend of longer reaction times for more memorized items, p = 0.0037, p = 0.06 (n.s.), and p = 0.00001, respectively.