2 ms·
Who says we're not "schizophrenic" now? Our common parlance attributes agency to blind chaotic forces in the universe; our visual systems are quasi-hallucinogen
by lukifer 8y ago
Who says we're not "schizophrenic" now? Our common parlance attributes agency to blind chaotic forces in the universe; our visual systems are quasi-hallucinogenic in their mechanisms (making all sorts of "optical illusion" mistakes); and while we like to think we're singular, holistically consistent beings, our actual behavior is often better described by nested (and competitive) sub-agent neuron clusters: https://meltingasphalt.com/neurons-gone-wild/ https://meltingasphalt.com/neurons-gone-wild/
It's worth considering that Jaynes' argument has as much to do with language as religion, positing a tightly coupled relationship between (spoken) language and subjective consciousness/qualia. "Gods" were in part an explanatory mechanism for the same phenomena of consciousness and (feeling of) decision-making that we experience today, which we now explain via "Self" narratives. In other words: rainbows, and our experience of them, did not fundamentally change, only our social stories about rainbows (God's Covenant -> Light Refraction).
Assuming one accepts a relationship between consciousness and language, it's also worth considering that language is a socially viral phenomenon, and so it makes sense that we might externalize consciousness explanations into a social space, via the common knowledge of our community's tribal gods, or a shared animist experience of nature. The innovation of the Self is a relatively recent one in this model; and even then, each of us frequently make the mistake of modeling other Selves after our own Self (the only god the secular world honors), resulting in biases like the Curse of Knowledge.
At any rate, Jaynes' argument is quite detailed and thorough, and difficult to make concisely. I found "Origins of Consciousness" to be a thought-provoking read, but I'm still not sure I'm willing to stake a strong claim For or Against the model.