Home > Papers > The Time-Course of Visual Categorizations: You Spot the Animal Faster than the Bird.

The Time-Course of Visual Categorizations: You Spot the Animal Faster than the Bird.

Macé MJ-M, Joubert OR, Nespoulous J-L, Fabre-Thorpe M, 2009
PLoS ONE 4(6): e5927. doi:10.1371/journal.pone.0005927

. . .[W]e compared human processing speed when categorizing natural scenes as containing either an animal (superordinate level), or a specific animal (bird or dog, basic level). Human subjects require an additional 40–65 ms to decide whether an animal is a bird or a dog and most errors are induced by non-target animals. Indeed, processing time is tightly linked with the type of non-targets objects. Without any exemplar of the same superordinate category to ignore, the basic level category is accessed as fast as the superordinate category, whereas the presence of animal non-targets induces both an increase in reaction time and a decrease in accuracy.

This is a fairly big claim and an interesting paper. The big wow here is that they show superordinate categories to be the same speed as basic level categories, in defiance of what is usually taken to be a defining characteristic of the ‘basic’ category.

First off, the study of the relative effect of the presence of superordinate vs unrelated competing items is an interesting one, and I’d like to know if there’s any more on it.

One thing that makes me a little wary is that they refer to categories such as ‘dog’ etc as inherently basic level. This notion, that some category has an inherent hierarchical role, is a hard one to put down. The real observation, though, is that there is some level, characterized by many interacting attributes (imageability, lexical simplicity, etc) that represents a semantic ‘sweet spot’, a cline of greatest density-per-level. There’s nothing making this fixed for all cognizers, and in fact one of the real game-changing ideas about this research is that it is expected to be relative. Put simply, there’s no reason to think that ‘basic’ for me is ‘basic’ for everyone, but that’s an assumption a lot of people make.

This same issue is revealed in the idea of an ‘entry level’ category different from a basic for atypical category members (penguin for bird, for example). The reification of the claddistic hierarchy interferes with understanding categorization. By a claddistic hierarchy, a penguin is a bird. But for many people, while they may cite this fact verbally, we might guess that their underlying categorization could be different. Positing an ‘entry level’ categorization to explain why penguins are categorized faster as ‘penguin’ than ‘bird’ could be succumbing to the ‘common knowledge’ that penguins are birds, and that therefore any other categorization is ‘wrong’.

Similarly, ‘dog’ and ‘bird’ are sibling children of ‘animal’ in a simple schematic hierarchy, but there’s no real reason to think that they relate this way in an individual’s experience of the world.

Another issue in this kind of experimentation (and not necessarily one the authors should have addressed) is modality. Pictures of birds and dogs, as used in this study, are probably better than line art used in others. But a big part of categorization of input experience of animate entities might be the way they move or their location relative to the observer – on the ground, in a tree, etc. Static images remove these cues and present a very different task for the brain. The idea is to get at ‘visual categorization’ alone, but that may be reifying an expectation belied by lots of embodied cognition evidence.

All this said, the overall claim, that a parallel representation of categorization is much more likely than a ‘two-step’ or hierarchical one, seems likely. They go on to point out that lexical activation of the basic-level terms could tilt the speed/accuracy odds and is compatible with a parallel-activation based approach. The Rosch et al model of categories, separate from the relativistic observations, is one that relies on a simple schematic hierarchy, which is predisposed to a linear understanding. An understanding based on a parallelized system, which the brain is, is likely to be a better one.

I’m not sure about the other claim, that abstract categories are inherently faster, for the reasons above and a few others (for example the object/context congruence that the authors themselves point out). But I do wonder whether what is linguistically basic may not always be what is experientially and cognitively basic for a given cognizer. Language is shaped by cognitive processes, but it is shaped slowly and grossly in some cases, and what is ‘basic’ for the speaker’s linguistic community may not always be for the speaker herself.

Categories: Papers Tags: , ,
  1. No comments yet.
  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: