
About Categorization
This is a series of blog posts about the learnings and discoveries I made while studying Cognitive Psychology at the University of Berkeley, California. It will introduce you to the basic concepts of human cognition; attention, memory, categorization, semantic organization, language, problem solving and decision making.
Part One
Let’s assume, that I am an Art Expert on a ‘hobbyist’ level and my friend just recently introduced me to this American artist, named Douglas Huebler. She showed me one of his early works from the 1970’s and described it as one of his most famous ‘art pieces’ — at least, that is the description and initial (category-) label my friend used.
Let us have a look at the object and describe it. It is a horizontal line, painted on paper and set in a wooden 76 x 56 cm frame. Underneath there is a statement: ‘The line above is rotating on its axis at a speed of one revolution each day.’ My friend asked me — ’Do you think this is art?’ And she added: ‘I could have done this.’ I answered, that perhaps she could do it, but that alone did not make her an artist. Even if she would leverage and apply the same skills, principles and rules the artist Huebler applied to his work, the outcome would still not be an ‘art’ object. At most, people would consider it a plagiarized work. Those were probably the first classifications I made in the conversation we had.
What was it that helped me to identify Huebler’s work as an ‘art object’ (the superordinate level)? To be more precise; what made it a conceptual art piece and not an abstract one or a traditional painting (the basic level)? I immediately ruled out the categorization strategies of the nearest-neighbor and average distance rule. There would have been too many items in a category that share similar attributes and even too many categories with similar items (black square, on paper, painted, framed, concrete and abstract, descriptive, etc.). The problem with some of the identified attributes is also, that they can have many continuous dimensions. The discussed art piece would be still the same, if you would change the colour of the line to green or change the line to a triangle. Or even if the line were painted onto a straight highway and made 300 meter long. It would be less ‘approachable’ (feature attribute), but it would still hold true to its ‘conceptual’ idea. The point is that you would have to deal with too many similar examples in a category, or with the problem of not finding a matching example at all.
I realize that this is counterintuitive to what Smith and Minda (1998) found in their research of expert learning. They both stated that experts often use exemplar models, since once they have memorized all the examples (“I’ve seen it — I know it all”), it allows them to quickly match a novel item with an item from a specific category — even on a subordinate level. This approach does not work with conceptual art, because every item, can be a new item and it requires a steady shift in perception and thinking.
I was able to observe, as Ashby and Maddox (2005) suggested, that we adopt our classification strategies depending on how categories (and/or objects) were constructed in the first place. Conceptual art was introduced in the late 60s and it must have thrown people off, since they could no longer apply their previously developed strategies of identifying art objects, which was mostly and/or purely based on prototype modelling. Now, you needed to understand the idea and concept, which made the execution or physical presence obsolete. A new attribute or feature that became part of the art object itself, was the transformation from art into language (Sol Lewitt, 1969 and Art & Language, 1967).
So the categorization strategy art experts should and probably do apply for the correct classification of conceptual art is a mix of conceptual, rule-based prototyping and contextually influenced feature frequency identification (both based on theories, both should focus on concepts and ideas). Going back to the example of Huebler’s work, these two strategies helped me to first, identify a ‘conceptual prototype’ while looking at theories, art manifestos and ideas that already existed in the realm of ‘concept art’ (Sol LeWitt, Sentences on Conceptual Art, 1969) and secondly, identify features that provide causal explanations to key questions. For example; does conceptual art need a physical presence? Well, the art object does not have to be executed by the artist itself. In fact, it does not have to be physically executed at all. Conceptual art is made to engage the mind of the viewer, rather than her/his eye or emotion. And so forth.
The beauty of art is that it leaves us with plenty of room for interpretation, which in return, makes it hard to classify its concrete meaning.
Part Two
My wife got me this beautiful guitar from Canada for my last birthday and I immediately wanted to learn how to play it (tip: Create a learning routine that fits into your daily life, otherwise, you will never pick it up.). On the first day, I practiced until my fingers hurt. I mostly looked up videos on Youtube and found three types of videos. The first category was concerned with teaching you how to learn notes (superordinate level) and how to apply or follow some rules of music composition. The second category dealt with teaching you the basics of playing a guitar — a string instrument (basic level). The third video category went straight into teaching you how to play specific songs on a guitar, as a beginner (subordinate level). My early conclusion was that learning to play a guitar is certainly both rule based and attribute based. If you want to play along, you not only need to follow the instructions but also need to follow the rules of the melody. If you want to play the right tone, you need to make sure you learned the attributes or features of a certain tone and finger-grip.
You might have guessed it — there were good and bad videos in all of the high level categories mentioned above, but how did I know which ones were which? I must have separated them into sub-categories. In order to qualitatively classify the material into good and bad videos, I first needed to consume many videos, which allowed me to form a prototype. This prototype was based on key feature attributes, that were stored in memory, such as; length of the video, language, speed of explanation, depth of information, visual cues, presentation of cues, sound quality etc. All these features had an effect on my ability/performance to learn from the material presented.
What comes along with learning an instrument is also the correct classification of sounds the instruments creates, when applying your newly acquired motor skills. The course textbook did not really go into the classification of sounds, tones and abstract melodies, but I assume that I applied some perceptual learning strategy that uses the auditory cortex and some prototype rule. I therefore needed to listen and store the basic tones first, and also create a spectrum for each single tone, which was influenced by the loudness, pitch and quality. Learning multiple, prototypical sounds and tones will also help me later, identify and categorize the instrument itself (musical instrument — string instrument — guitar — etc.). With sufficient practice, I will start automating the recognition of tones and eventually switch to a faster categorization strategy and make use of exemplar classification. I know when I am there, when the ‘sound is stuck in my ears’.

