A Structure from Manipulation for Text-Graphic Images

What is the difference between pattern structure 1 and pattern structure 2 for the same spatial arrangement of identical visual atoms?

Well, first let me cop a plea      I make no strong claims about the psychological reality of pattern structures *.

Instead, I make some armchair predictions. If people were presented with these simple visual objects:

Then ..

1. If we asked people to draw the objects on a piece of paper, they would be unlikey to draw the objects in the order of pattern structure 2      ie in the order As-box, B, A, Bs-box. Rather, they would be more likey to draw them in the order of pattern structure 1      ie As-box, A, Bs-box, B (or possibly A, As-box, B, Bs-box; that's fine too, see pattern structure 3 below).

2. If we presented these objects on a computer screen and asked people which objects were grouped with what other objects      say for the purpose of dragging, ie the old "what moves with what" question      then again the prediction is that pattern structure 1 would be the more likely response.

3. Likewise if we presented these objects and asked people to construct them in some graphics editor that supported grouping, pattern structure 1 would be the more likely response.

... so I call these various groupings "Structures from Manipulation" and use the fine tree lines (spider webs) to denote that structure.

Further workings of the notation:

1. Sometimes pattern order may be important, and sometimes not, in which case patterns are just sets. Thus the difference between pattern structure 1 and pattern sturcture 3 can be ignored for many purposes.

2. And, the notation can represent each structural variation.

3. Text graphic object notation is meant as a tool for thinking about the structure of diagrams, including the empirical investigation of how people actually structure such text graphic objects.

4. There is also a computer implementation of the notation within a graphics editor, with facilities for processing text graphics and running experiments.

5. The implementation supports only strict trees, i.e. an object can be a member of only one pattern. However, the notation can show an object with multiple parents, which might be useful for certain diagrams, as in the case below.

* The lisp heritage of tg obj notation leaves it open to the dangers of such claims, cf. the infamous "psychological reality of lisp" conference.

  Back to first page