Tokens and bullae are fascinating pieces of early human economics, but the core of Schmandt-Besserat's theory -- the idea that certain tokens developed into carvings of words and concepts -- is widely discounted by archaeologists today.

For one example of why, consider the earliest depiction of the word for sheep: a circle with a cross. Sheep were one of the largest commodities in Mesopotamian trade, and yet only a handful of tokens matching that crossed-disk shape have ever been discovered. If transactions involving sheep were taking place daily, they should be among the most plentiful of tokens. These token systems are an ingenious human invention, but they should be seen as the precursors to mathematical accounting, not writing.

But didn't most writing systems evolve from accounting systems?

Chinese writing is pretty likely to have come from deviation (dropping turtle shells into fires and interpreting the cracks).

Not sure of the origin of mesoamerican writing.

Middle eastern writing's origin (motivation) is also unclear.