Perceptual grouping refers to the human visual ability to extract significant image relations from lower-level primitive image features without any knowledge of the image content and group them to obtain meaningful higher-level structure,

see Perceptual organization and visual recognition & What makes viewpoint invariant properties perceptually salient?: A computational perspective, & Perceptual organization of two-dimensional patterns & Human and Machine Vision, Computing Perceptual Organization...

take a look at this: http://www.lsl.nie.edu.sg/icce2012/wp-content/uploads/2012/12/C1-f-221.pdf

Eyes are focused by attention, but attention can be driven by conscious and sub-conscious processes - maybe read Attention and Effort by Daniel Kaheman (it's available online as a pdf). Gesalt principles allows system 1 (i.e. sub-conscious) processes to effectively object-ise / 'group' / generalise the world into manageable chunks. Any conflict to the associative model of mental schemas causes a discourse that has to be explained - which requires system 2 (conscious) processes. Accordingly, design that follows Gesalt principles is easy to process (as you don't have to think about it). Design that plays with or conflicts with Gesalt principles is in conflict with the brains understanding of the world, and will draw the focus of attention - hence causing a movement the eye. Such design also causes huge cognitive load resulting in cognitive overload - which in turn can be disastrous (thought possibly measurable via pupil size !?). Yarbus did some very interesting work on eye-tracking focus and task (with different types of viewing) in 1967 - but it is still relevant (can also be found online).

Also, it seems like you would need an experiment where the proximity of the visual stimuli is randomized across participants. That will untangle the perceptual grouping that you are after from the proximity effect. Interesting idea!

Although it is not directly related to Gestalt principles, you might find it helpful to look at the multiple object tracking literature that used eye tracking. Attentionally tracking multiple objects is to some extend supported by grouping the tracked objects into an higher order object ( see http://pbs.jhu.edu/research/Yantis/publications/PDF/Yantis-CogPsych-1992.pdf ). And the eye-tracking research within this area treated gaze towards the centroid of this higher order object as evidence for grouping:

http://www.sciencedirect.com/science/article/pii/S0010027707002776

http://www.tandfonline.com/doi/abs/10.1080/13506285.2010.495878#.Us6zrbS8boY

http://www.journalofvision.org/content/10/4/19.full

http://www.psychology.sunysb.edu/gzelinsky-/publications/Z&N%20(2008).pdf

There is a couple of studies showing very different pursuit eye movements when the observers group moving lines and when they see them as separate entities, which could be related to Franks’ response:

Stone, Beutter and Lorenceau (1999) ‘Visual motion integration for perception and pursuit’ in Perception, http://www.perceptionweb.com/perception/fulltext/p29/p2979.pdf and Beutter and Stone (2000) http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=40141

You can see what the coherent (grouped) and incoherent motion looks like here: http://web.mit.edu/persci/demos/Motion&Form/demos/square_lines/square_lines.html

Hi,

I'm not an expert on this, but last year I came across a paper that showed a very clear relationship between eye position and the corresponding percept of a Necker cube (Einhaüser et al, 2004, Are switches in perception of the Necker cube related

to eye position? European Journal of Neuroscience, Vol. 20, pp. 2811–2818). That makes me think that one possibility could be to use some bistable or multistable gruping configuration and try to relate each possible percept with the spontaneous eye movements.

Hope it helps.

Article Are switches in perception of the Necker cube related to eye position?

The first stages of perceptual grouping has to take place in the eye periphery previous to focal attention. So the first stages of perceptual grouping is done in the periphery at low spatial resolution and is also very sensity to movement of the objects, The first stages of perceptual grouping do not classify or recognize the specificity of the objects but only likely retina locations where objects of the world of particular interest might be located. By fixating such group, the visual system can dedicated the full resource of the visual system for actually classifying the type of object. Once that is done, the attentional system may divert the fovea to other location without losing the identity of the object but just indexing its locations though peripheral processing.

The gestalt laws of perceptual organization are very crude description of how perceptual grouping is actually done. The actual grouping is done by hiearchical crease structure detection in the scale-space-time of the images. In layman terms, an grey scale image on the retina is like a topographical luminance surface and its analysis is done at different spatial and temporal scales. What is most important in such topographical surfaces are the networks of ridges and valleys linking the critical points of the surface. Objects whose images are projected on the retina have surfaces whose luminance surfaces have almost the same topological structures and the crease structure at different spatial and temporal scales capture the essential feathure of these structure. The analysis is carried out from the coarse to fine spatial scales.

For more details, see : Ph.D. thesis: The perception of the Image World

http://www.collectionscanada.gc.ca/obj/s4/f2/dsk1/tape8/PQDD_0023/NQ51844.pdf

You might also want to look at the work of Eileen Kowler on eye movements and attention

Dear Giorgio, Stephen, and Louis, thank you for further clarifying the concepts about perceputal grouping, Gestalt laws, and eye movements. These clear definitions are very helpful to reconfigure the question and to breakdown the large question into smaller, manageble pieces. Dear Dr. Inman. Frank, David, Pilar, and Elizabeth, thank you all for your specific suggestions and good examples about the existing studies on this topic. Your answers give me very concrete starting points. I will look into the papers you guys mentioned to get a better grasp. I really appreciate everybody's time and thought in movin this question one step forward. : -) Best wishes, Sarina

You could maybe found some answers in Anirvan S Nandy & Bosco S Tjan (2012). Nature Neuroscience. They deal mainly with crowding, but also with grouping and anyway they are related...

Thanks, Lukasz.. That's a good suggestion. This article you mentioned is very interesting. I found it.

http://www.nature.com/neuro/journal/v15/n3/full/nn.3021.html

Thanks, Sarina

Hello Sarina,

You have been well served with theory, so i thought that I would take a more pragmatic, nuts and bolts approach to what I think is your problem. Am I right in thinking that you want to present visual stimuli to observers, record their spontaneous eye movements and fixations on these stimuli, and then, by analysing the distribution of eye-fixations, fixation times and perhaps scan paths, find evidence for or against what you call “grouping”? Let’s leave what we mean by “grouping” for the moment. When you have recorded the x-y coordinates of your observers’ fixations on your stimuli, their fixation durations, it is possible to look at the distribution of fixations and duration as a three-dimensional map for just one observer or cumulatively across all your observers. You can do this easily using a three-dimensional chart in Microsoft Excel, and see in detail where any observer has “grouped” his or her fixations and fixation duration on any one visual stimulus. You can also look at the accumulation of fixations and fixation times across one or many observers and/or across one or many stimuli.

Some may argue that this method of judging where “grouping” of fixations and duration on the stimuli has occurred is just a bit too subjective. Some people may see groupings in the distributions that others do not. One step towards a more objective analysis of the distributions is to use cluster analysis. There are at least a couple of cluster analyses that can be used with eye-movement data - k-means and the Wallace-Boulton Information measure. I have attached a paper explaining how to carry out the analyses described above. There are other papers on the use of cluster analysis on eye-movement data, but until we are sure that this approach is going to be useful to you, we can leave these and also treatment of scan paths aside.

There are also two ways to approach your problem. You could begin with a theory that predicts where “grouping” will occur. For example, if your stimuli were alphabetic characters, you may predict that attention would be concentrated (grouped) around the most discriminating features of the characters. You may have reason to believe that some parts of your visual stimuli are more “salient” than other parts, and will attract more attention. The described data-analysis methods can help you test these theories. A second approach is simply to have no theory and look at where attention is grouped on your stimuli, and on the basis of your results, work towards some predictive theory that could be tested on new stimuli. Note too that the distribution of attention is going to be strongly influenced by the task demands, the nature of the stimuli and instructions to the observers.

It will not have escaped your notice that I have moved from talking about eye movements and fixations to “attention” - a controversial move that could be discussed. I have also introduced the term “salience” - a real can of worms that one! And, while I understand the Gestaltist notion of grouping, I am a little unsure of what you and the other commentators mean by the same term.

I have attached the paper referred to in my previous answer.

Dear Cyrill, I'm thrilled by your answers. Thank you! What you described in the first paragraph is exactly what I hope to do. The more objective method of doing cluster analysis on the distribution is a great idea, too. In fact, my motivatipn behind this quesiton is to further understand the development of young infant's perceptual organization. Up to date, several Gestalt grouping principles have been investigated in infants with the (total) looking time measure. For exmaple, grouping by proximity and by lightness similarity seem to be functioning at about 3 months (Quinn et al 2006; Quinn, Burke, and Rush, 1993) ; while grouping by form similarity develops around 6-7 months (Quinn et al 2002). The evidence is primarily based on whether infants show a significantly longer (overall) looking to a particular pattern to "infer" if grouping occurs or not. I think if we have a way to look further at the microstructure and distribution of infant's eye movement on the visual sitmuli, we shall be able to make more progress in this line of work. As for the infleunce of attention, I guess infants' attention is more bottom-up or data-driven in nature, as compared to adults'. Thus, infants' perceptual grouping shall be largely determined by stimulus strength (or saliency if you want to 'open the can of worms" : ) ). Any idea on that? Well, I really appreciate your wonderful answer and feedback on my thoughts. Thanks for adding the paper (cyrill00002.pdf), too!

Let me attached a few papers on the development of Gestalt principles:

One more interesting paper by Quinn et al.

The best work I know on this topic is by Mary Peterson in a number of research areas. One of her papers is "Illusory concomitant motion in ambiguous stereograms" JEP:HPP, 1986, 12(1), 50-60. You might also look at the edited volume by Peterson, Gilliam, and Sedgwick "In the Mind's Eye: Julian Hocherg on the perception of pictures, films, and the world."

Hi Dr. Cutting, Thank you for joining the discussion and pointing out to this line of work by Dr. Peterson and the classic book about Dr. Julian Hochberg's constructive view on perception. Thank you! I will look into that. Best wishes, Sarina

 Hi Sarina,

I just came across this discussion.  We are going to be using Eye tracking in a Perceptual organization task (DEPOT) with schizophrenics and I was wondering if you have any results or further knowledge about research in this area since this discussion first came up.  Thanks