If you think about that, 2x3 or 3x2 is the practical minimum. A fully empty glyph and a single dot wouldn't be really useful, and any translation would be too a

If you think about that, 2x3 or 3x2 is the practical minimum. A fully empty glyph and a single dot wouldn't be really useful, and any translation would be too ambiguous for 2x2 grid, so we only have 8 possibilities: .@ .@ @. @. @@ @@ @@ @@ @. @@ .@ @. .. .@ @. @@ This is too small for any actual, natural written language. 2x3 and 3x2 significantly increase possibilities and it is indeed not too hard to pick unique glyphs for a large subset of alphabets. Rotokas [1] only uses 12 unaccented Latin alphabets for example, so the following would marginally work: .@ @@ @@ .@ @. @@ @@ .. @. @@ .. .. @@ @. .@ .@ @@ @@ @@ @@ @@ .@ @@ @@ @@ @@ @@ .@ @@ @@ @. @. .@ .@ @@ @. A E g I K O P r S T u v Maybe you can make the question much more interesting by allowing non-square and/or non-uniform pixels, i.e. segmented displays. That still has a lot of design potentials even to this day. [1] https://en.wikipedia.org/wiki/Rotokas_language https://en.wikipedia.org/wiki/Rotokas_language