Exp Brain Res (2004) 159: 400 404 DOI 10.1007/s00221-004-2111-y RESEARCH NOTES Giovanni Galfano. Elena Betta. Massimo Turatto Inhibition of return in microsaccades Received: 26 May 2004 / Accepted: 25 August 2004 / Published online: 8 October 2004 # Springer-Verlag 2004 Abstract Inhibition of return (IOR) is the term used to describe the phenomenon whereby stimuli appearing at recently attended locations are reacted to less efficiently than stimuli appearing at locations that have not yet been attended. In the present study, we employed a typical IOR paradigm with peripheral uninformative cues while participants maintained their eyes at fixation. Eye position was monitored at a high sampling rate (500 Hz) in order to detect miniature eye movements called microsaccades, which have been shown to be crucial for avoiding disappearance of visual image. However, recent studies have demonstrated a close relationship between covert endogenous attentional shifts and the direction of microsaccades. Here, we demonstrate that the direction of microsaccades can be biased away from the peripheral location occupied by a salient, although task-irrelevant, visual signal. Because microsaccades are known not to be under conscious control, our results suggest strong links between IOR and unconscious oculomotor programming. Keywords Inhibition of return. Microsaccades. Visual perception. Oculomotor programming Introduction The ability to effectively explore the environment is important for both humans and other species. To this end, the visual system is equipped with hard-wired mechanisms that enhance the processing of salient stimuli at attended locations and withdraw attention away from previously attended stimuli. The mechanism that seems to mediate this latter process is known as inhibition of return (IOR), and its primary function might lie in promoting the G. Galfano. E. Betta. M. Turatto (*) Department of Cognitive Science and Education, University of Trento, Via Matteo del Ben, 5, 38068 - Rovereto, Italy e-mail: massimo.turatto@unitn.it Tel.: +39-0464-483529 inspection of novel objects (Posner and Cohen 1984). A tight link between IOR and the programming of saccadic eye movements has been established through a variety of research techniques including behavioral studies in both adult humans (Rafal et al. 1989) and newborns (Simion et al. 1995), as well as neuropsychological (Sapir et al. 1999) and neuroimaging studies (Lepsien and Pollmann 2002). Usually, the oculomotor involvement in IOR has been established by monitoring saccadic eye movements (Abrams and Dobkin 1994; Watanabe 2001). The purpose of the present study was to address any possible links between IOR and microsaccades, which are miniature eye movements executed without awareness during fixation (Martinez-Conde et al. 2004). Specifically, we hypothesized that IOR could modulate microsaccade direction, as recent studies have shown that the orientation of microsaccades is correlated with orientation of visual attention (Hafed and Clark 2002; Engbert and Kliegl 2003). Engbert and Kliegl (2003) have produced a coherent set of experiments consistently showing that microsaccade direction, in a standard spatial cueing paradigm, was correlated with the direction of endogenous shifts of attention. Such shifts of attention were induced by highly predictive (80% validity) symbolic cues presented at fixation, indicating the likely location for appearance of the target. In line with these findings, Hafed and Clark (2002) showed a similar microsaccadic pattern in a more complex paradigm using peripheral cues that were always informative with respect to target location. However, recently, Rolfs et al. (2004) reported microsaccade orientation counter to cue direction in response to informative peripheral cues. Here, we also demonstrate that microsaccade direction can be biased away from the peripheral location occupied by a salient, although task-irrelevant, visual signal in a standard IOR paradigm. Note that in the Hafed and Clark (2002) and the Rolfs et al. (2004) studies, both the endogenous and the exogenous components were likely at work, as the authors used peripheral cues (thus engaging exogenous control) that were predictive of the location of target appearance (thus strongly engaging endogenous
401 control). By contrast, in our experiment the exogenous component of attentional control was thought to occur in isolation. The stimulus onset asynchrony (SOA) we incorporated in the present experiment was chosen in order to tap onto IOR. Previous studies with similar experimental settings have proved that a 1,000-ms SOA is indeed able to produce IOR in simple detection tasks (e.g., Berlucchi et al. 1989; Klein 2000). Materials and methods Participants The study population comprised 13 undergraduate students (ten females and three males, mean age 28.3 years). Informed consent to participation was obtained from all participants, in accordance with the Declaration of Helsinki. All subjects reported having normal or corrected-to-normal vision. None of them reported having any neurological disease. All were naïve as to the purpose of the study. Apparatus, stimuli, and procedure The experiment took place in a dimly lit room. Eye movement data were acquired and recorded via an Eye- Link II system (SR-Research, Ontario, Canada) with a sampling rate of 500 Hz and a spatial resolution of less than 0.005 of visual angle. Response times (RTs) and accuracy were also recorded. Timing of events and generation of stimuli were controlled by a custom-made C program running under Windows 2000 on a Pentium IV Dell PC. The stimuli were presented on a CRT Dell 19- inch color monitor (1,280 1,024, 75 Hz). The monitor was placed at eye level in front of the participants. Participants were seated with the head positioned on a chin rest, so that the distance between the eyes and the monitor was 48 cm. At the beginning of the experiment, a standard nine-point-grid calibration was performed. The background was black and stimuli appeared in white. Participants were shown a fixation spot (0.73 in diameter) centered on the screen, which lasted throughout the trial, along with two white boxes (whose side measured 3 ) aligned on the horizontal meridian and equidistant from fixation (12.4 on the left and the right). At the beginning of the fixation frame, a warning tone (800 Hz) was delivered for 50 ms in order to inform participants that the drift correction phase was about to begin. This phase began 250 ms after the tone had been delivered. Correct fixation was accomplished when gaze position was detected in a region four times as large as the fixation spot. If this could not be detected after 800 ms, the display went black and the calibration phase was repeated. After the drift correction had been completed, there was a fixed 1,000-ms pre-cue baseline phase. No changes took place in the display during either the drift correction or the pre-cue baseline phase. Then the cue was delivered, which consisted of one of the two boxes becoming thicker and brighter for 50 ms. After a fixed 950-ms interstimulus interval (ISI), resulting in a 1,000-ms SOA, the target (a bright square whose side measured 0.7 ) was turned on for 50 ms centered on either the cued or the uncued box with the same probability. Participants were allowed 1,000 ms to respond by pressing, with their dominant hand, a button of a joypad connected to the computer. RTs were registered from target onset. The key-press ended eye movement recording, which was restarted at the beginning of the fixation frame of the next trial. The intertrial interval was set at 2,000 ms. During this period, the display went black and no eye movement recording took place. The experiment consisted of 224 trials in total, divided into seven blocks of 32 trials each. Half were catch trials (i.e., no target was presented). Half of the remaining trials (56) were valid (the target appeared at the cued location), while the other half (56) were invalid (the target appeared at the uncued location). Participants were instructed to press the button as fast as possible on target-present trials and to withdraw from responding on catch trials. Also, they were allowed short breaks between blocks. Participants were explicitly informed that the cue was not informative of the location at which the target would appear, and were instructed to maintain fixation throughout a trial. The feedback for misses was a 1000-Hz tone delivered for 500 ms. Microsaccade detection Microsaccades were investigated monocularly, as the vast majority of microsaccades are conjugated (Møller et al. 2002). Also, we restricted our analyses to horizontal microsaccades. We used a modified version of the algorithm described by Engbert and Kliegl (2003), adjusted to our higher sampling rate (see also Engbert and Kliegl 2004). Eye positions exceeding 1 from the centre of the screen were discarded. The time series of eye velocities was then computed as a moving average of velocities over nine gaze positions. Independent velocity thresholds for the horizontal and vertical components were set to six standard deviations of velocity. Microsaccades were detected as outliers in the velocity space according to these thresholds. To further remove noise, only microsaccades with a minimum duration of six samples (12 ms) were analyzed. Microsaccades with an amplitude exceeding 1 were also discarded. Before analyzing the data, we verified that the eye movements we classified as microsaccades satisfied the velocity amplitude relationship criterion (Zuber et al. 1965). Figure 1a shows the peak velocities of microsaccades from the whole set of participants as a function of their amplitudes. The plot contains 8,623 microsaccades. The correlation coefficient was very high (r=0.8), and the regression slope k of the v=ka function (v= peak velocity, A= amplitude) was (73.9±0.3) s 1, a value that is
402 consistent with those reported by Møller et al. (2002). Therefore, the validity of the microsaccade detection algorithm was supported by the velocity amplitude relationship. Results RTs Responses on catch trials were rare (less than 5%) and were not analyzed. A t-test with type of trial (valid vs invalid) as factor was conducted on mean RTs for correct responses to check whether the classic behavioral signature of IOR was present. It revealed that participants responded significantly slower on valid (mean=410 ms, SD=189) than on invalid trials (mean=395 ms, SD=194), t (12)= 2,505, p<0.05, thus showing a reliable IOR. Eye movement data Microsaccade rate An overall rate of 1.2 microsaccades per second, averaged over all trials and participants, was found in the ±1,000 ms interval centered on cue onset. The evolution of the microsaccade rate (number per second) was then computed by using a moving time window of 100 ms centered at the current point in time. The microsaccade rate was averaged over all trials (including catch trials) and all participants. As shown in Fig. 1b, the onset of the peripheral cue elicited a sort of microsaccadic inhibition lasting for about 200 ms. After this decrease, microsaccade rate increased to a maximum at about 300 ms after cue onset. Microsaccade direction First, we selected 15 200-ms time windows, ranging from 1,000 ms before to 2,000 ms after cue onset. Note that, because eye movement recording was ended by key-press upon target appearance, data in the time windows ranging from 1,000 to 2,000 ms referred to cue onset mainly represent microsaccades executed during catch trials. The logic we adopted for analyzing the data was the same as that employed by Hafed and Clark (2002). We computed the differences between the frequency of microsaccades towards and opposite to the cue and then performed a series of multiple comparisons (two-tailed t tests) with a zero value (i.e., random noise) for each of the 15 selected time windows in order to check whether a bias, at any time windows, could be observed. To minimize the possibility of making a type II error while controlling type I error, we adjusted the α level for the number of comparisons following the False Discovery Rate procedure (Benjamini and Hochberg 1995). A significant positive difference from zero indexes a bias towards the cue location, whereas a negative difference points to a bias in the direction opposite to the cue (Fig. 2). The only comparison reaching a significant difference was that concerning the time window centered 300 ms after cue onset, t(12)= 5,372, p<0.0001, showing a significant tendency of participants to execute a microsaccade in the direction opposite to the cue. No other effects were significant either at the pre-cue baseline or after cue onset. It is worth noting that the delay in the directional bias against the cue is strikingly similar to the SOA at which evidence for IOR usually emerges in RTs for simple detection tasks (Klein 2000). Fig. 1 a Peak velocities of microsaccades as a function of amplitude. The plot contains 8,623 microsaccades from the whole pool of participants. b Absolute microsaccade rate as a function of time. Data from target-present trials and catch trials are collapsed. Negative values on the abscissa refer to the pre-cue baseline, while zero refers to cue onset Discussion It is well established that IOR has profound consequences not only on the temporal features of saccades, but also on their metrics (Godijn and Theeuwes 2002; Ro et al. 2000; Watanabe 2001). In the present paper we have extended
403 such results by showing that even if participants are maintaining their gaze at fixation, an oculomotor signature for IOR can be found when microsaccade direction is examined. This evidence fits well with the notion of a tight link between attentional orienting and eye movement programming (e.g., Corbetta 1998; Rizzolatti et al. 1987). Because IOR is mediated through the involvement of subcortical oculomotor pathways (Sapir et al. 1999), and because microsaccades cannot be evoked voluntarily (Martinez-Conde et al. 2004), our data are in agreement with the notion that the neural circuitry controlling microsaccades includes the superior colliculus. As shown in Fig. 2, no statistically significant bias emerged for microsaccades in direction of the cue shortly after its presentation. A remarkably similar pattern has recently been reported by Rolfs et al. (2004), who employed peripheral informative cues and observed microsaccades opposite to the cued location with a peak around 300 ms after cue onset. Yet, if microsaccades represent an oculomotor correlate of covert attentional orienting (e.g., Engbert and Kliegl 2003), these results might be surprising. Although many studies have shown that exogenous orienting (i.e., with peripheral uninformative cues) usually produces a biphasic pattern in performance, with a short-lived period of enhanced processing in valid trials followed by a much longer period of impaired processing (see Klein 2000), other studies have shown that this is not always the case (e.g., Tassinari and Campara 1996; Maruff et al. 1999). We only used a long SOA, and therefore we do not have an RT measure to check whether an advantage in performance emerges for valid trials when the target is presented in close temporal proximity with the cue. However, even if one accepts the view that attentional capture should be accompanied by an oculomotor capture (involving saccades in overt orienting and microsaccades in covert orienting), there are at least two possible explanations for the lack of directional bias towards the cue immediately after its onset. On the one hand, one might simply argue that no bias was observed in the cue direction because as soon as a change in the image takes place (e.g., cue onset in the present study), microsaccades are suppressed, as shown by reduction in the microsaccade rate following cue onset (also see Engbert and Kliegl 2003). This is because one of the functions of microsaccades is to prevent the retinal image from fading (e.g., Engbert and Kliegl 2004; Leopold and Logothetis 1998; Martinez-Conde et al. 2000). The implication would be that microsaccades cannot be observed shortly after a change in the pattern of retinal illumination has occurred because these tiny eye movements do not take place until retinal receptors have undergone a certain amount of adaptation (see, however, Coppola and Purves 1996; Horwitz and Albright 2003). Thus, if a shift of attention occurs within the period of microsaccade suppression, possible links, if any, between microsaccade direction and attentional allocation would be simply undetectable. Fig. 2 Microsaccade rate computed as a function of the difference between the frequency of microsaccades towards and opposite to the cue. The continuous line represents microsaccade rate computed every 2 ms (see text for details). Histograms represent the mean microsaccade rate in 200-ms windows. Bars represent 95% confidence intervals. Zero on the abscissa refers to cue onset. Positive frequency values index a bias towards the cue, whereas negative values index a bias in the opposite direction. The only peak reaching a significant difference from zero was the one centered 300 ms after cue onset. The plot contains the data from the whole pool of participants On the other hand, one could reasonably assume that participants were aware that the peripheral but unpredictive visual transient was potentially able to grab their attention, and that such spatial capture might have evoked an involuntary saccade (producing an oculomotor capture). Because of this, and because participants were clearly instructed to keep their gaze at fixation, they might have exerted a strong form of oculomotor control on eye position. As a result, any eye movement towards the cue would have been inhibited. This was also Rolfs et al. s (2004) explanation of their results. In their study, cues were predictive; therefore it is not clear whether the endogenous component due to cue validity or the exogenous event per se was responsible for the inhibition. In the present study, peripheral cues were not informative. Thus, our data demonstrate that such inhibition occurs irrespective of such endogenous demand. There is also a further possibility. After being directed towards the cue, attention would have been shifted away (possibly towards the opposite location) as predicted by IOR, which is also in accordance with the RT pattern that emerged in our experiment. Importantly, this reorienting of attention to the unexplored location might have taken place without awareness. In this context, evidence that attention can be oriented unconsciously is available from both normal adults (e.g., McCormick 1997) and neurological patients (Danziger et al. 1998). If this hypothesis were correct, then it might be speculated that inhibition of microsaccades opposite to the cue did not take place because participants were not aware they were orienting their attention to the corresponding location. Note that this
404 interpretation is, at least partially, challenged by the results reported by Rolfs et al. (2004). In fact, they observed that peripheral informative cues induced a microsaccade bias opposite to their location. Clearly, because, unlike in the present study, the cue was predictive of the most likely location for the target to appear, participants had no reason to remove their attention from the cued location. In such a context, microsaccade bias opposite to the cue could hardly have resulted from a shift of attention. Future studies will investigate this issue in more detail. A final topic that should be addressed is whether there are any possible implications of the present results for the mechanisms governing IOR. Different views have been expressed as to the mechanisms that trigger IOR (i.e., the causes of IOR), some emphasizing the role of attentional factors and others focusing on motor components (see Klein 2000). In this regard, it is important to remember that the present study was aimed at exploring the effects of IOR on microsaccade orientation rather than at providing any causal explanation about the mechanisms underlying IOR. However, our data are consistent with both the oculomotor suppression hypothesis (e.g., Tassinari and Campara 1996) and the oculomotor priming hypothesis (Rafal et al. 1989). According to the first hypothesis, IOR would simply result from participants suppressing an exogenous saccade towards the cue because of the instructions to maintain fixation. On the other hand, the oculomotor priming hypothesis argues that saccade preparation (either exogenous or endogenous) is both necessary and sufficient to elicit IOR. Although our data are clearly in accordance with the oculomotor suppression hypothesis, because the peripheral cue feasibly triggered the preparation of a saccade towards its location, oculomotor priming can also account for our microsaccade results, as both accounts of IOR are based on the strong involvement of oculomotor programming. Acknowledgements This research was supported by an MIUR grant to M. Turatto. We are indebted to Massimiliano Pastore for helpful advice on data analysis, and to Reinhold Kliegl for allowing us to use the microsaccade detection algorithm and for very useful comments on previous drafts of the manuscript. We also thank Martin Rolfs and an anonymous reviewer for valuable suggestions. References Abrams RA, Dobkin RS (1994) The gap effect and inhibition of return: interactive effects on eye movement latencies. Exp Brain Res 98:483 487 Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc B 57:289 300 Berlucchi G, Tassinari G, Marzi CA, Di Stefano M (1989) Spatial distribution of the inhibitory effect of peripheral non-informative cues on simple reaction time to non-fixated visual targets. Neuropsychologia 27:201 221 Coppola D, Purves D (1996) The extraordinarily rapid disappearance of entoptic images. Proc Natl Acad Sci U S A 93:8001 8004 Corbetta M (1998) Frontoparietal cortical networks for directing attention and the eye to visual locations: identical, independent, or overlapping neural systems? Proc Natl Acad Sci U S A 95:831 838 Danziger S, Kingstone A, Rafal RD (1998) Orienting to extinguished signals in hemispatial neglect. Psychol Sci 9:119 123 Engbert R, Kliegl R (2003) Microsaccades uncover the orientation of covert attention. Vision Res 43:1035 1045 Engbert R, Kliegl R (2004) Microsaccades keep the eyes balance during fixation. Psychol Sci 15:431 436 Godijn R, Theeuwes J (2002) Oculomotor capture and inhibition of return: evidence for an oculomotor suppression account of IOR. Psychol Res 66:234 246 Hafed ZM, Clark JJ (2002) Microsaccades as an overt measure of covert attention shifts. Vision Res 42:2533 2545 Horwitz GD, Albright TD (2003) Short-latency fixational saccades induced by luminance increments. J Neurophysiol 90:1333 1339 Klein RM (2000) Inhibition of return. Trends Cogn Sci 4:138 147 Leopold DA, Logothetis NK (1998) Microsaccades differentially modulate neural activity in the striate and extrastriate visual cortex. Exp Brain Res 123:341 345 Lepsien J, Pollmann S (2002) Covert reorienting and inhibition of return: an event-related fmri study. J Cogn Neurosci 14:127 144 Martinez-Conde S, Macknik SL, Hubel DH (2000) Microsaccadic eye movements and firing of single cells in the striate cortex of macaque monkeys. Nat Neurosci 3:251 258 Martinez-Conde S, Macknik SL, Hubel DH (2004) The role of fixational eye movements in visual perception. Nat Rev Neurosci 5:229 240 Maruff P, Yucel M, Danckert J, Stuart G, Currie J (1999) Facilitation and inhibition arising from the exogenous orienting of covert attention depends on the temporal properties of spatial cues and targets. Neuropsychologia 37:731 744 McCormick PA (1997) Orienting attention without awareness. J Exp Psychol Hum Percept Perform 23:168 180 Møller F, Laursen ML, Tygesen J, Sjølie AK (2002) Binocular quantification and characterization of microsaccades. Graefe s Arch Clin Exp Ophthalmol 240:765 770 Posner MI, Cohen Y (1984) Components of visual orienting. In Bouma H, Bouwhuis DG (eds) Attention and performance X. Erlbaum, Hillsdale, NJ, pp 531 556 Rafal RD, Calabresi PA, Brennan CW, Sciolto TK (1989) Saccade preparation inhibits reorienting to recently attended locations. J Exp Psychol Hum Percept Perform 15:673 685 Rizzolatti G, Riggio L, Dascola I, Umiltà C (1987) Reorienting attention across the horizontal and vertical meridians: evidence in favor of a premotor theory of attention. Neuropsychologia 25:31 40 Ro T, Pratt J, Rafal RD (2000) Inhibition of return in saccadic eye movements. Exp Brain Res 130:264 268 Rolfs M, Engbert R, Kliegl R (2004) Microsaccade orientation supports attentional enhancement opposite a peripheral cue. Psychol Sci 15:705 707 Sapir A, Soroker N, Berger A, Henik A (1999) Inhibition of return in spatial attention: direct evidence for collicular generation. Nat Neurosci 2:1053 1054. Simion F, Valenza E, Umiltà C, Dalla Barba B (1995) Inhibition of return in newborns is temporo-nasal asymmetrical. Infant Behav Dev 18:189 194 Tassinari G, Campara D (1996) Consequences of covert orienting to non-informative stimuli of different modalities: a unitary mechanism? Neuropsychologia 34:235 245 Watanabe K (2001) Inhibition of return in averaging saccades. Exp Brain Res 138:330 342 Zuber BL, Stark L, Cook G (1965) Microsaccades and the velocity amplitude relationship for saccadic eye movements. Science 150:1459 1460