Uncorrected Proof The Neural Basis of Parallel Saccade Programming: A Functional Imaging (fMRI) Study Yanbo Hu and Robin Walker Abstract ■ The neural basis of parallel saccade programming was exam- ined in an event-related functional imaging (MRI) study using a variation of the double-step saccade paradigm. Two double- step conditions were used: one enabled the second saccade to be partially programmed in parallel with the first saccade while in a second condition both saccades had to be prepared serially. The intersaccadic interval, observed in the parallel pro- gramming (PP) condition, was significantly reduced compared with latency in the serial programming (SP) condition and also to the latency of single saccades in control conditions. The fMRI analysis revealed greater activity (BOLD response) in the frontal and parietal eye fields for the PP condition compared with the SP double-step condition and when compared with the single- saccade control conditions. By contrast, activity in the supple- mentary eye fields was greater for the double-step condition than the single-step condition but did not distinguish between the PP and SP requirements. The role of the frontal eye fields in PP may be related to the advanced temporal preparation and increased salience of the second saccade goal that may mediate activity in other downstream structures, such as the superior colliculus. The parietal lobes may be involved in the preparation for spatial remapping, which is required in double- step conditions. The supplementary eye fields appear to have a more general role in planning saccade sequences that may be related to error monitoring and the control over the sequence of responses. ■ INTRODUCTION Saccades are made to shift gaze and attention onto ob- jects of interest enabling detailed analysis of the visual scene. As we can only shift our eyes to one object at any one time, there has been a natural tendency to re- gard saccade programming as a serial process, but some behavioral studies have provided evidence showing that the saccadic system can program more than one saccade in parallel. Becker and Jürgens (1979) showed that sac- cades made to targets that moved in two steps could be separated by very short intersaccadic intervals (ISIs) that were much shorter than the time required to gener- ate a single saccade. Short ISIs have also been observed in other situations such as the antisaccade task (Hallett, 1978) and in visual search (Theeuwes & Godijn, 2002; Findlay, Brown, & Gilchrist, 2001; Theeuwes, Kramer, Hahn, & Irwin, 1998). In these paradigms, incorrect saccade can be followed, after a very short ISI, by a secondary correc- tive saccade directed to the saccade goal. Importantly, the ISI separating the first and second saccades is much less than the time to generate a single response (c.f. McPeek, Skavenski, & Nakayama, 2000; Mokler & Fischer, 1999; Amador, Schlag-Rey, & Schlag, 1998; Weber, Dürr, & Fischer, 1998; Hooge & Erkelens, 1996; Viviani & Swensson, 1982). The short ISI period has been interpreted as showing that the second corrective saccade was programmed in parallel (pipelined) with the first erroneous response. Walker and McSorley (2006) used a variation of the double-step saccade paradigm to investigate parallel pro- gramming (PP) without having to rely on an examination of the smaller proportion of double responses made on error trials (cf. Sheliga, Brown, & Miles, 2002). Partici- pants made a first stimulus-elicited saccade, was followed by a second (“voluntary”) saccade made to a goal indicated by an arrow cue. A robust reduction in second saccade la- tency was observed compared with that of comparable sin- gle saccades. A similar reduction in second saccade latency (ISI) was also found when a first “voluntary” saccade, made to a location specified by an arrow cue, was followed by a second saccade made to a peripheral target. First, saccade latency was found to be modulated by the required direc- tion of the second saccade, which is consistent with the view that both saccades may be programmed on a com- mon “motor map” (Godijn & Theeuwes, 2002) such as that formed by neurons in the intermediate layers of the supe- rior colliculus (SC; McPeek & Keller, 2002). The Neural Basis of Parallel Saccade Programming Single-neuron studies have demonstrated a role of the SC in the PP of saccades (McPeek, Han, & Keller, 2003; McPeek & Keller, 2002). McPeek and Keller (2002) revealed that neural activity associated with a second corrective University of London © Massachusetts Institute of Technology Journal of Cognitive Neuroscience X:Y, pp. 1–12