Something has been bugging me about my use of the term “word illusion.” Though I know how I mean to use it and why, it always seems a bit awkward and roundabout. I’ve figured out why.
It is our ordinary use and understanding that is (somewhat) illusory. When we see words on the page, we see and experience them as words-in-full, not as mere signifiers/word-forms. We don’t even experience words in a foreign language as mere signifiers/word-forms. Rather, we experience them as words we don’t understand. They aren’t word-forms, they’re merely words.
And the same is even more true for spoken language. We hear and understand the meanings. The word-forms come before consciousness only if something goes wrong. And then we experience them as words-gone-wrong.
When I talk of the word illusion, then, I’m talking of taking our ordinary understanding of and experience of words into a domain where that understanding can get us into trouble. In the case of literary criticism it leads us to interpret all over the damned place. In NLP it leads us to misinterpret or over interpret what the machine is doing.
And I suppose why I was a bit taken aback when I read Michael Gavin’s fascinating article, “Vector Semantics, William Empson, and the Study of Ambiguity.” For he also referred to vector semantics as computational semantics, to which my immediate response was, “No no no, that’s not computational semantics. Computational semantics is different.” By that I meant that computational semantics (proper) is what I’d studied in graduate school, Old School symbolic constitutive semantics. Vector semantics is fine, but it is a very different beast from, well, constitutive semantics – a term I coined to contrast with vector semantics. People need to know the difference.
But enough of that, which is in a way a diversion. What set me off this morning is thinking back to the days when I was fascinated by dictionaries and encyclopedias. I’d guess it was middle school.
I’d look up a word in the dictionary, read its definition, and still be a bit puzzled. I found the usage examples particularly useful – that’s why they’re there, no? Often as not the definition itself would contain a word or two I didn’t understand, so I’d go look them up. And the same thing would happen. And so I’d happily loop my way through the dictionary reading word after word after word. And that’s certainly how I understood and experienced those words, as words-in-full, even though each dictionary and information about pronunciation and possibly alternative spellings. They were simply other attributes of the word.
I’d do the same thing with the encyclopedia. Of course, I didn’t loop through those entries quite so rapidly. And no doubt I went back and forth between the encyclopedia and the dictionary.
This was several years before I went to college and tried to puzzle out signifier, signified, and thing or referent. How would I have reacted to that three-way distinction back then when I was looping through the dictionary and the encyclopedia?
 Michael Gavin, Vector Semantics, William Empson, and the Study of Ambiguity, Critical Inquiry 44 (Summer 2018) 641-673: https://www.journals.uchicago.edu/doi/abs/10.1086/698174 Ungated PDF: http://modelingliteraryhistory.org/wp-content/uploads/2018/06/Gavin-2018-Vector-Semantics.pdf.