We read with great interest the recent PNAS paper by Hanning and colleagues, in which they use a variant of Craighero et al’s Eye Abduction paradigm to test the proposal that exogenous covert attention is constrained by the oculomotor range (Smith & Schenk 2012; Casteau & Smith 2019). If you’re not familiar with the paradigm, figure A illustrates the setup. The idea is that by abducting the eye by ~35-40 degrees it is possible to present stimuli at locations that can be seen but cannot be reached with a saccadic eye-movement. This neat procedure decouples perception and oculomotor control, and in 2012 we showed that peripheral cues beyond the oculomotor range did not summon attention (Smith, Rorden & Schenk 2012).
In the Hanning version of the task participants are shown 4 channels of noise, 3 within the range of eye-movements and one beyond. One of the locations is cued for 50ms, and 100ms later a probe appears in one of the channels for 50ms. Participants must report the orientation of the probe with button press. People are really bad at this task, and accuracy is only above chance at the cued location. Critically, this attentional facilitation for the cued location is observed even when the cued channel is beyond the range of eye-movements, contrary to our previous findings. Hanning and colleagues argue this is problematic for premotor theory, and argue the data are more consistent with selection-for-action models like VAM. They conclude that covert, exogenous orienting of attention is not constrained by the range of eye-movements, contrary to our previous findings.
The paper is beautifully presented, the experiment is cleverly designed and on first inspection these data look clear cut. There can be no doubt that the peripheral cues are producing enhanced target detection. So, do we need to abandon our position that covert, exogenous orienting is dependent on oculomotor activation? Perhaps not yet. Hannings’ data are convincing, but as they note in their paper, there are considerable differences in the stimuli used in our lab (e.g. Smith et al., 2012, Casteau & Smith 2018), by Craighero et al., (2012) and by Michalczyk et al., (2018) that show an effect of eye-abduction on covert attention, and those used by Hanning and colleagues. One potentially crucial difference is the difficulty of the task. Hannings’ task is extremely difficult. In fact, it is impossible to accurately discriminate the target unless it is attended (See fig 2b: dprime is ~0 at all uncued locations – you can see the task for yourself here). To understand why this is a problem we need to first consider the rationale for using a non-predictive peripheral cueing task. The reason typically given is that a non-predictive cue gives the participant no incentive to endogenously attend to the cued location, so allows a ‘pure’ measure of exogenous attention. This rationale makes sense when the task is simple detection task or a speeded discrimination, because the participant detects the target regardless of whether or not it appears at the cued location, and the differences emerge in the speed of target detection. However, this rationale does not hold true when the task is so difficult that the target is never detected at uncued locations. Why? Because under these conditions the cue is highly informative from the perspective of the participant; they only ever perceive a target when they attend a cued location, and never detect targets at uncued locations (a similar issue is found in the change blindness literature- see Smith & Schenk 2010). It therefore makes complete sense for participants to adopt an attentional set of ‘endogenously attend cued location’ in order to correctly perform the task. We know that endogenous shifts of attention can be observed at short SOAs, for example in Muller & Rabbit (1989) seminal study there is a beneficial effect of endogenous attention at 100ms SOA, and Tipples (2002) has shown a similar effect for arrows. It therefore seems likely that the peripheral cue used by Hanning et al., engaged both endogenous and exogenous attention, and it was the endogenous component of attention that produced the attentional benefit. If participants are endogenously attending the cued location, the results are in fact in exact accordance with our previous findings and those of Klein and colleagues (1980;1994), which were that endogenous orienting is independent of oculomotor control. A less critical, but still noteworthy issue is that the sample is rather small (n=6 in the critical covert attention task). In contrast, papers showing impaired exogenous attention beyond the oculomotor range tend to have larger samples (N= ~20-25), so there is also a risk that this study was somewhat underpowered to detect the interaction of interest. Hanning et al., also make the valid point that previous studies using eye-abduction did not establish the range of eye-movements for their participants. However, in our 2018 paper we measured each participants own oculomotor range, and again showed a selective deficit for covert exogenous attention at locations beyond the range of eye-movements when the eye was centred in its orbit .
To briefly summarize, Hanning et al., show that a peripheral cue produces attentional facilitation beyond the EOMR. The evidence for this effect is very clear, and we do not dispute that participants did indeed covertly orient to the cued location. However, we do dispute the claim that this shift of attention is exogenous in nature. Rather, we believe it is more likely that participants endogenously oriented to the peripheral cue because it was the only way to successful complete the task, and that the improved performance at the cued location is caused by this endogenous shift of attention. We agree with the conclusion that the data are problematic for premotor theory, but only because they show that endogenous orienting is possible to locations that cannot become the goal of a saccade.
We asked Hanning & colleagues if they would like to reply to our comment, and will add it here ASAP if they do. Nina will be joining us in Durham for our workshop and we’re looking forward to some lively discussions! (there are still some places if you fancy joining us)
References
- Casteau, S.; Smith, D.T. Covert attention beyond the range of eye-movements: Evidence for a dissociation between exogenous and endogenous orienting. Cortex [Google Scholar] [CrossRef]
- Casteau, S & Smith, D.T. (2019). Associations and Dissociations between Oculomotor Readiness and Covert Attention. Vision 3: 17.
- Craighero, L.; Nascimben, M.; Fadiga, L. Eye position affects orienting of visuospatial attention. Biol.2004, 14, 331–333. [Google Scholar] [CrossRef]
- Hanning, N., Szinte, M,. & Deubel H. Visual attention is not limited to the oculomotor range. Proceedings of the National Academy of Sciences 2019, 116 (19) 9665-9670; DOI: 10.1073/pnas.1813465116
- Klein, R.M. Does oculomotor readiness mediate cognitive control of visual attention? Perform. 1980, 8, 259–276. [Google Scholar]
- Klein, R.M.; Pontefract, A. Does Oculomotor Readiness Mediate Cognitive Control of Visual Attention? Revisited! In Attention and performance XV: Conscious and Nonconscious Information Processing; Umiltà, C., Moscovitch, M., Eds.; The MIT Press: Cambrige, MA, USA, 1994; pp. 333–350. [Google Scholar]
- Smith, D.T.; Schenk, T. The Premotor theory of attention: Time to move on? Neuropsychologia 2012, 50, 1104–1114. [Google Scholar] [CrossRef]
- Michalczyk, L.; Paszulewicz, J.; Bielas, J.; Wolski, P. Is saccade preparation required for inhibition of return (IOR)? Lett. 2018, 665, 13–17. [Google Scholar] [CrossRef] [PubMed]
- Smith DT & Schenk, T. (2010). Inhibition of return exaggerates change blindness. Quarterly Journal Of Experimental Psychology 63(11): 2231-2238.
- Smith, D.T. Smith, D.T.; Schenk, T.; Rorden, C. Saccade preparation is required for exogenous attention but not endogenous attention or IOR. Exp. Psychol Hum. Percept. Perform. 2012, 38, 1438–1447. [Google Scholar] [CrossRef] [PubMed]
Daniel T. Smith & Soazig Casteau,
Dept of Psychology, Durham University UK
ty. Now, imagery isn’t our usual sort of thing, but the chance to attend a free conference in our own backyard seemed to good to miss. On day 1 we saw some fascinating presentations from students in Shaun Boe’s lab who had travelled all the way from Dalhousie in Canada. The first thing to note is that most of the contemporary research was focused on AOMI (which everyone pronounced AY-OH-ME, much to Soazigs irritation). AOMI involves action observation (i.e. watching an expert do the task) while simultaneously engaging in mental imagery (imagining performing the action). The key message I took from these talks was that although it was generally agreed that mental imagery was jolly helpful for improving skills, there was no real agreement about either the optimal ‘dosage’ or at what point during skill acquisition AOMI was most effective. Boe’s group showed some pretty nice evidence that AOMI was most effective during skill acquisition; the ‘cognitive’ and ‘associative’ phases in Fitts & Posner (1967) model of expertise. In a particularly interesting finding, Sarah Kraeutner argued that imagery could reduce endpoint variability without necessarily improving overall accuracy in a dart throwing task. However, a number of audience members commented that although the dart-throwing paradigm is a classic task, it has problems as a model of skill acquisition, particularly with respect to controlling prior experience of the task. Also in this session was a fascinating talk from Jack Binks on using
imagery to rehabilitate motor skills after stoke. Richard Ramsay then gave an eclectic keynote covering cognitive and brain mechanisms of social perception, which I found particularly interesting with respect to our work with Geoff Cole on perspective taking and 
rt to reconcile or integrate these views, and we agreed this could be an area that is ripe for further study. The day 2 keynote was given by Robert Hardwick (KU Leuven), who presented a fascinating meta-analysis of studies examining imagery, action observation and action execution, and finished with an appeal to develop more theoretically principled models of mental imagery. Again, I was struck by the parallels between the dominance of Simulation Theory in the motor imagery literature and that of OMRH/PMT in the literature on spatial attention, and the need to develop beyond these ideas. There was also time for a bit of light schmoozing around the posters, and I was delighted to talk briefly with Zoe Franklin about TMS and eyetracking, Elaine Poliakoff about her work on imagery in Parkinsons Disease, Shaun Boe about his ideas on the neural and cognitive mechanisms underpinned AOMI, Gabriel Valadez Roque about using imagery to rehabilitate stroke patients in Mexico and Martin Edwards, with whom I mainly exchanged academic gossip 😉.
