3/17/2015

How expertise changes the way you see

When you know a lot about something, it doesn't just change how you think--it also changes how you see.

Stock image courtesy of RGBStock.

A long tradition in psychology, education, and business involves comparing people with little experience in an area to others with some amount of expertise (e.g., novice chess players vs. grandmasters, medical students vs. doctors).  Despite this two-group approach to studying it, expertise is actually a continuum, with "experts" in studies ranging from typical competent members of a profession (e.g., doctors) to world-class performers (e.g., chess grandmasters). When "experts" are trained in the lab to recognize artificial creatures1, the level of expertise is lower still--as few would volunteer to spend the famous "10,000 hours of practice"2 in a lab learning to categorize such items. Yet, even this sort of "expert" both thinks about and sees artificial creatures differently than people who have never seen them before. They haven't just memorized a set of images, either: this change in perception extends to members of the category that they have never seen before.

This shift in perception that comes with expert knowledge is called perceptual expertise. It is highly category-specific. For example, a car expert might display perceptual expertise when viewing modern cars but not antique ones, or vice versa3. Perceptual expertise is not limited to vision, but it has been studied less in other senses, so I will only talk about vision here.

Perceptual expertise affects many aspects of vision, including...

1) Memory for Visual Patterns
Many areas of expertise involve learning to recognize complex visual patterns. Chess is a great example. A chess grandmaster can look at an arrangement of pieces on the board and tell you who is attacking and who defending, what each player's likely next move will be, and perhaps even what famous games involved this arrangement. (Some arrangements, called "positions," are named). They are seeing not just the placement of each piece, but a meaningful whole defined by the relationship between them.

A series of classic studies by Chase, Simon, and Gobet demonstrated this elegantly by asking chess masters and ordinary players to remember the positions of a large number of pieces. These positions were either taken from actual master-level chess games, or were completely random (and often would never occur in an actual game). This experiment is shown in part B of the image below:
The image comes from this description of Chase, Simon, and Gobet's chess studies.

If chess experts are learning the meanings of configurations of pieces, then they should have an advantage over the novice players for real game positions, but not for random positions. And that's exactly what the studies found.

The research team also tested the working memory spans of both groups, and these were similar. So chess masters don't remember arrangements of chess pieces better than novices because they're better at keeping in mind large amounts of arbitrary information in general. Instead, they are able to group pieces with particular configurations into a meaningful "chunk" which can then be recalled as a unit, instead of 15 separate pieces.

2) Focus on the Whole Over the Parts
In chess, experts learn to recognize configurations across a set of objects. However, some domains of expertise lead experts to recognize configurations of parts within one object. This phenomenon was first studied in face perception (in which most people can be considered experts). However, it has since been noted in several other areas of expertise, such as cars. It should be found for any objects where the arrangement of parts is important for recognizing an object, but not for objects where specific features, like color or texture, are the defining ones4.

How do we know people are focusing on the relationship between an object's parts rather than the parts themselves? One way is to turn the object upside down. This disrupts the spatial relationship between the parts more than it disrupts the look of the parts themselves. As a result, experts are slower to recognize a specific object upside-down than they are when the same object is right-side-up. This disruption is called the "inversion effect."

Rossion and Curran4 examined the "inversion effect" in car experts, by showing them pictures of faces and cars that were either upright or inverted. Car expertise was determined not only by self-report, but by the ability to match 112 images of cars based on model. The faces served as a comparison condition. We know that inversion affects faces; the question was whether for car experts, inversion would also affect cars, and to a similar extent.

Above: face and car stimuli from Rossion & Curran (2010).

Interestingly, for novices, who have extensive familiarity with faces but less with cars, inversion reduced accuracy for both faces and cars. This is different than you might expect if expertise alone is responsible for inversion effects. However, it makes sense when you consider that even novices have often viewed cars right-side-up, but never upside-down.

For novices, inversion affected faces more than cars, both in terms of accuracy and reaction time. For experts, inversion reduced reaction time for both faces and cars similarly, and in fact, it non-significantly hurt performance for cars a bit more. Furthermore, the degree of expertise (ability to accurately match car models) correlated with the degree to which inversion slowed car identification.

This correlation is important. It makes it less likely that some random factor is responsible for the differences between experts and novices in car perception.  Given that expertise is a continuous phenomenon, it's also much more convincing to find that the amount of expertise correlates with the size of the inversion effect.

In short, just as most people learn to recognize upright faces by their configuration of features, so car experts had learned to recognize upright cars by theirs.

Other studies have found similar inversion effects with various objects of expertise, including artificial objects5, fingerprints6, chess boards7, and words8 . As in all areas of research, some studies have not found any significant effects9, perhaps because the categories of stimuli used were broader than the categories of objects participants were actually expert in. While these studies were part of a debate about whether faces are objects of extreme expertise or innately "special", that's not what makes them interesting. Whatever you think about faces, what matters is that developing expertise with recognizing other specific kinds of objects leads you to view them as a whole, not just as the sum of their parts.

3) Detecting and Recognizing Objects
We all know that people are good at recognizing objects that are important to them. Objects of expertise are important to experts, so it won't surprise you that they notice them better. For example, bird experts notice birds that to anyone else, might seem to blend into the background. (For a beautifully-written meditation on how expertise transforms the look of an ordinary city block, see On Looking: Eleven Walks with Expert Eyes by Alexandra Horowitz).

In the lab, you can show people a large number of objects, including objects of expertise and other common objects.  On some trials, you ask people to identify their object of expertise and on others, to pick out some other sort of object. Experts are faster and more accurate at finding objects of expertise than other sorts of objects. Novices find all sorts of objects equally well (all other things being equal).

Below is an example of this task from a study by Hershler and Hochstein10. Experts were car or bird experts, and control images were either faces or "other objects" (miscellaneous).


Unfortunately, I don't know of studies that investigate visual search in a naturalistic context, like a living room scene or a beach scene. The problem with arrays of objects like this one is that it's difficult to control variables like color, texture, and brightness for all objects or all arrays. But such variables can make some objects easier to find than others, regardless of what type of object they are. It gets even more complicated when you include some experts in living objects (like birds) and others in nonliving objects (like cars), since these likely differ in their low-level visual properties.  For this reason, I don't use (and don't plan to use) this sort of visual search task in my own research.

Other Perceptual Changes?
In the interests of completeness, I should note that a further, controversial set of studies on video games has suggested that some types of expertise may improve people's ability to quickly spread attention across large areas of space or multiple objects. Because these studies are less well-accepted, though, I will save discussion of them for a future post. However, I can safely conclude that the perceptual changes I've discussed in this post are the best established but not the only ones. We will likely discover more as we investigate further areas of expertise.

What we know...and what we still need to figure out
From birdwatching to chess, developing expertise seems to change how you see.

This basic principle seems to hold true across areas of expertise. However, different areas of expertise seem to impose different perceptual demands. For example, chess appears to improve perception of relationships between objects, leading to better pattern recognition, while car expertise improves perception of relationships between features within an object, leading to better object recognition.

Research groups tend to focus on one area of expertise in isolation, and then draw conclusions from their group of experts about expertise in general.  For expert knowledge, this works fine. The conclusions about problem-solving Michelene Chi drew about physics experts11 and young dinosaur experts12 seem to be accepted in domains as diverse as medical diagnosis and business leadership, as even a cursory Google search will indicate. However, perceptual expertise seems to be more domain-specific. Yet I know of no systematic examination of which areas of expertise will change perception in what ways. This is something I might work on during my PhD, so rather than speculate further, I'll simply note that this is an important problem to solve. Ideally, it would be possible to determine how any area of expertise changes how people see.

References

  1. Isabel Gauthier, Pepper Williams,  Michael J. Tarr, and James Tanaka (1998). Training "greeble" experts: A framework for studying expert object recognition processes. Vision Research 38:15-16, 2401-2428.
  2. K. Anders Ericsson, Ralf T. Krampe, and Clemens Tesch-Romer (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review vol. 100 iss. 3, 363-406. http://psycnet.apa.org/index.cfm?fa=search.displayRecord&uid=1993-40718-001 
  3. Cindy M. Bukach, W. Stewart Phillips, and Isabel Gauthier (2010). Limits of generalization between categories and implications for theories of category specificity. Attention, Perception, & Psychophysics vol. 72, iss. 7, 1865-1874.
  4. Bruno Rossion and Tim Curran (2010). Visual expertise with pictures of cars correlates with RT magnitude of the car inversion effect. Perception vol. 39, pp. 173-183.
  5. Isabel Gauthier and Michael J. Tarr (1997). Becoming a "Greeble" expert: Exploring mechanisms for face recognition. Vision Research vol. 37, iss. 12, 1673-1682. Open access! http://www.sciencedirect.com/science/article/pii/S0042698996002866
  6. Thomas A. Busey & John R. Vanderkolk (2005). Behavioral and electrophysiological evidence for configural processing in fingerprint experts. Vision Research vol. 45 iss. 4, 431-448. Open access paper: http://www.sciencedirect.com/science/article/pii/S0042698904004365  
  7. Merim Bilalic, Robert Langner, Rolf Ulrich, and Wolfgang Grodd (2011). Many faces of expertise: Fusiform face area in chess experts and nonvices. Journal of Neuroscience vol. 31 iss. 28, 10206-10214.
  8. Chien-Hui Kao, Der-Yow Chen, and Chien-Chung Chen (2010). The inversion effect in visual word form processing.  Cortex vol. 46 iss. 2, 217-230. http://www.sciencedirect.com/science/article/pii/S0010945209001427
  9. Rachel Robbins & Elinor McKone (2007). No face-like processing for objects-of-expertise in three behavioral tasks. Cognition vol. 103, 34-79. https://www.msu.edu/course/psy/802/altmann/802/Ch2-3-RobbinsMcKone07.pdf 
  10. Orit Hershler and Shaul Hochstein (2009). The importance of being expert: Top-down attentional control in visual search with photographs. Attention, Perception, & Psychophysics vol. 71 issue 7, pp. 1478-1486.
  11. Michelene T.H. Chi, Paul J. Feltovich, and Robert Glaser (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science vol. 5 iss. 2 121-152. http://onlinelibrary.wiley.com/doi/10.1207/s15516709cog0502_2/pdf 
  12. Camilla Gobbo and Michelene Chi (1986). How knowledge is structured and used by expert and novice children. Cognitive Development vol. 1 iss. 3, 221-237. http://www.sciencedirect.com/science/article/pii/S0885201486800028