What is counted is length
The Zipfian distribution describes the tendency for units that occur often in a communication system to be short and rare ones to be long. In human languages it is a well-documented pattern, usually read as speech shaping itself around economy. The team led by the University of Edinburgh carried that measure over to bird and whale sound: the Bengalese finch, the humpback whale, and humans as a comparison.[1]
The method was borrowed from human infants. Infants work out word boundaries by tracking which sounds occur together; the team applied the same approach to birdsong data. The finches' vocalisations are organised into statistically coherent note sequences, and those sequences follow the same Zipfian pattern as human speech. The pattern emerges in the early learning stage the birds pass through while listening to adults. Humpback whale song shows a comparable structure.[1]
The line the researchers draw
With findings of this kind the sentence usually slides one step further and animals are said to speak. The team draws the line itself: language is found only in humans, though we share the planet with other species whose communication systems evolve culturally. What is measured is the distribution of unit lengths within a system; meaning, grammar and reference stay outside the measurement.[1]
That the same statistic appears in three species at once points to a shared pressure: a system passed on by learning across generations erodes toward shapes that are easy to learn, and the units used most often get shorter. There is an explanation standing against that. Short sounds may simply be repeated more because they are cheap to produce, in which case the source of the pattern is the cost of production itself and no cultural learning is required. The write-up does not choose between the two.[1]
Three species, and how wide a claim?
The breadth of the comparison sets the breadth of the claim. The three species in the Edinburgh team's hands — a finch, a humpback whale and a human — are spread widely across living things, yet three points do not draw a curve. Since the species that acquire song by learning are not limited to these, how widespread the pattern is stays open for now.[1]
The comparison that would really test the claim follows from this. Measure the same length pattern in a species whose song is innate rather than culturally learned: if it is absent, the transmission explanation gains ground; if it is present, the explanation has to be sought in the cost of production. The question of how humans are alone then becomes narrower and more answerable: does the solitude sit in the statistics that organise the signals, or in what we do with them?[1]