What waggle dance is
Honeybee foragers returning to the hive from a profitable food source perform a stereotyped figure-eight dance on the vertical surface of a honeycomb. The dance has a 'waggle' segment in the middle where the bee shakes its body while moving in a straight line; the direction of the waggle relative to gravity encodes the direction of the food source relative to the sun's azimuth, and the duration of the waggle encodes the distance to the food. Other bees in the hive observe the dance and depart for the indicated location. Foragers update their dance to account for the sun's continuing motion across the sky between dance performances. This is not folklore; this is the demonstrated behavior, replicated thousands of times across multiple bee species.
The proof came in three steps.
How von Frisch demonstrated it
The proof came in three steps. First, careful observation established that returning foragers reliably performed the figure-eight dance. Second, controlled experiments showed that bees observing the dance subsequently flew to the location the dance encoded — not random locations, not the location the dancing bee had come from regardless of dance content, but specifically the location the dance's direction-and-duration parameters specified. Third, manipulation experiments (dance interruptions, dance in non-standard orientations, dance with controlled angle deviations) showed that altering the dance's encoded parameters altered the recipient bees' destinations in predicted ways. The combined evidence was unusually clean: senders encode information, receivers decode information, the encoding-decoding relationship is reproducible and quantifiable.
Why this is the gold standard
Three properties together. First, the sender-side encoding is rigorous: waggle direction and duration map quantitatively to spatial parameters in measurable ways. Second, the receiver-side decoding is demonstrated: observing bees fly to the predicted destinations, not random ones. Third, the manipulation experiments confirm that the encoding-decoding relationship is causal: altering the encoded parameters alters the decoded behavior in the predicted direction. Most claimed examples of animal communication 'decoding' fail at least one of these three tests. The waggle dance passes all three with high confidence. That's the gold standard, and very few other communication systems in any species approach it.
What it doesn't include
Waggle dance encodes spatial coordinates (direction and distance to a food source) plus rough quality information (vigor of the dance correlates with food-source desirability). It does not encode the type of food, the identity of the discoverer, anything about who shouldn't go, or anything that would normally count as linguistic content. Calling waggle dance a 'language' overreaches what's encoded. Honeybee communication is referentially shallow — it points to coordinates rather than naming objects — but within that shallow scope, it is rigorously decoded. The closest fair characterization is 'a quantitative spatial map projection,' not 'a language.'
Why nothing else has matched
Subsequent attempts to find animal communication systems with comparable encoding-decoding clarity have mostly produced weaker evidence. Vervet monkey alarm calls (Cheney and Seyfarth) come close on the encoding side; the receiver-side evidence is good but less quantitatively rigorous than the bee work. Bottlenose dolphin signature whistles (Janik and others) show individual-identity encoding but the equivalent of waggle-dance's direction-and-distance quantitative mapping doesn't exist. Crow individual signatures and behavioral-context clusters are descriptively well-characterized but the rigorous receiver-side experiments comparable to von Frisch's manipulation work haven't been done. The waggle dance bar is high; nothing in vertebrate animal communication has cleanly matched it.
What this means for AI animal-language research
The fundamental challenge AI animal-language research faces is reproducing waggle-dance-level clarity in species where the communication system is more complex and the experimental constraints are tighter. Bees are easy to experiment with — small, social, hive-bound, relatively unconstrained by welfare considerations. Vertebrates — especially long-lived intelligent social vertebrates like crows or cetaceans — present ethical and methodological constraints that make waggle-dance-style manipulation experiments very hard. The likely future of animal-language AI doesn't include another von Frisch demonstration for crows or whales; it includes accumulating descriptive evidence that gradually constrains what the communication systems could plausibly be doing, without ever achieving the same gold-standard clarity.