Dolphins Don't Use Sentences: How Badly Executed Research has Created a False Narrative of Cetacean Communication
The media has been all atwitter this last week about a rediscovered paper from 2011 that “proves” that “dolphins speak to each other in complete sentences.” It’s garnered international awareness as a story… which is really awkward, considering that the paper that they’re referencing (Ryabov, 2016 - a republication of the original article from 2011) makes massive jumps in logic to draw appealing but very over-enthusiastic conclusions. Upon a careful read-through, Ryabov appears to have identified differences in patterns of the clicking noises that dolphins produce and re-described them in a new way - equating these units of sound with phonemes in spoken human language. The problem is that’s a pretty huge and unscientific leap to take in response to a single study that only deals with a 34 second-long recording of two animals. That’s a single recording - not enough to base a new theory on, and certainly not enough to justify where Ryabov goes from there. He immediately concludes in the discussion of the 26 paper that because we’re seeing something that ‘might’ look like phonemes, an assumption can be made as to how dolphins convey information by combining them (there is exactly zero data to support combinations of noncoherent pulse calls having any greater information content than each singularly, and in fact no evidence that single noncoherent pulses have any information content at all - all Ryabov has proved is that they exist as discrete units of produced sound). This makes the current media furor incredibly misleading because not only is there not enough definitive evidence to support any sort of new scientific theory, the original data has been so over-interpreted that the original paper is full of potentially erroneous and definitely unproven statements. So,let’s break down what we know about dolphin communication so far, what information and statements the paper actually contains, how the media are misinterpreting the contents.
Dolphin communication as we know it:
Most dolphin species have a number of different types of communication that they utilize - body language, tactile cues, tonal sounds (whistles, chirps, screams), and pulses sounds called clicks or bursts (a click is a single sound - bursts are sets of clicks produced in very quick succession). Tonal communication in dolphins has been pretty well studied - it appears that most dolphins have a signature whistle that identifies them to their pod, and most groups have some sort of shared repertoire of stereotyped (fixed) set of whistle calls that communicate some types of concrete information. Pulse calls aren’t as well understood, but they’re divided into two categories of use: echolocation and communication. Single clicks are used by dolphins for echolocation, but burst pulses (also called click-trains) are sets of clicks produced too quickly to provide any useful information on the environment; researchers have noted a strong correlation between burst pulse production and states of excitement/aggravation/general emotional arousal. Click-trains which are are characterized by “regular, incremental changes in energy, frequency and inter-pulse intervals” are called coherent pulses, whereas click-trains where the energy level, speed, and time between pulses vary irregularly are called noncoherent pulses.
Before we can look at Ryabov’s paper(s) on communication and the assumptions he makes about having found language, it’s important to define what language is and what the current scientific understanding is regarding the potential ability for dolphins to possess it. I’m going to borrow a beautifully elegant explanation from the Dolphin Communication Project to get us started:
“Human language is a system of combining small meaningful elements into larger elements, forming phonemes, words and sentences that allow humans to convey infinitely complex amounts of information. (...) scientists now think that some animals may possess small parts of this system within their own communication systems that allow them to generate some basic forms of human language-like communication, but nearly everyone agrees that only humans are able to learn and use language to the complicated extent that we know as ‘language’. (Dolphin Communication Project)”
In some studies, dolphins have proven that they’re able to learn to understand basic “artificial language” such as hand signs, tactile computer interfaces, or artificially generated whistles, no study to date has found evidence of grammar or strings of complex concepts that would imply an innate dolphin “language.” Instead, we’ve ascertained that specific stereotyped whistles probably have specific meanings - but they don’t seem to be utilized as parts of a larger communicative message.
These ‘small meaningful elements’ are what Ryabov thinks he has found in the noncoherent pulse calls of bottlenose dolphins. And indeed, he may have - he’s identified what appear to be discrete units of sound. However, the problem is the fact that the conclusions he draws from this data - let me remind you again, from a single clip that is 34 seconds long - by far overreach what is reasonable in this case.
Ryabov’s ‘groundbreaking’ study:
As we go through the study, I’m going to differentiate between things stated in the original 2011 publication and the 2016 re-publication. This is important because the same information is represented in a very different way - and sometimes not even included - in the more recent version of the article.
The 2016 article on his study starts with a stated assumption that dolphins have a spoken language, which isn’t really how to do good science. A well thought-out scientific inquiry would start with asking what we’d need to look for in communications to determine if dolphins do or do not have language, rather than starting off looking to “reliably measure and analyze the noncoherent pulses as the most likely acoustic signals of the hypothetic spoken language of the dolphins.” In contrast, the 2011 version of the paper stated that “the studying of characteristics and functions of the dolphins’ acoustical signals is the purpose of the study.” As both publications are reporting on exactly the same acoustical data, it’s obvious the 2016 paper was edited in order to convey the author’s chosen message instead of the vaguer but scientifically appropriate investigative hypothesis of the 2011 original.
The paper basically states that both dolphins involved in the study produced sets of noncoherent pulses in 10 ‘packs’ during the course of the 34-second recording. These packs were differentiated from each other due to the fact that there was a much longer interval of time between packs than there was between the calls within a pack. The dolphins were not trained for the experiment, and were simply recorded when at rest of their own accord. So far, so good. The 2016 paper skips directly to discussion that draws conclusions about interactions patterns without exposition of the data leading to those conclusions; the 2011 publication contains much more information about how the data was analyzed.
The first incorrect claim: dolphins didn’t interrupt each other.
The most important information from the 2011 publication that is left out of the 2016 paper actually contradicts one of the major statements in the paper to support dolphin “conversations.” The 2016 discussion section of the paper states that “the dolphins took turns in producing pulse packs and did not interrupt each other”, assuming that meant that they “listened to the other (...) before producing [their] own.” This is the statement from which both the author and the media have decided that this vocalization exchange is the equivalent of a human conversation. However, that’s not actually what happened in the study as described by the original 2011 publication. According to the english version of the study I read, the dolphins did not overlap each other for whistles or coherent calls - but they did overlap the critical noncoherent calls. (It’s important to note that dolphins are capable of producing whistles, coherent calls and noncoherent calls simultaneously and independently of each other). To quote the paper:
“In the pack (2) both dolphins produced pulses almost simultaneously and therefore, it is possible to assume, that they not understood each other, therefore Yasha has conceded and has allowed the lady to express oneself (packs of 3 - 5). In the pack (6) again, both dolphins almost simultaneously produced pulses, and though Yasha gradually increased their level, from the pulse to the pulse, Yana, apparently, again not listened him, then he loudly shout (last pulse in the pack 6). Then, after the short pause Yasha produced the packs of pulses (7 - 10) already up to the end of the record, having lowered the loudness of voice (the packs of 8 - 10).”
Oddly heteronormative assumptions about dolphin behavior and somewhat garbled english aside, it’s clear from the 2011 article that these dolphins certainly did produce noncoherent calls overtop of each other. When looking at the exact sequence that occurred during the noncoherent calling as called out by Ryabov and the data (Yasha & Yana vocalize 5s into the recording -> Yasha stops vocalizing around 8s -> Yana continues vocalizing through approximately 11s -> Yasha & Yana vocalize simultaneously at 14s with Yasha’s volume increasing -> Yasha vocalizes intermittently from 18s - 28s) in absence of any other behavioral information, it’s not even possible to discern if they’re responding to the timing of each other’s calls, or if they’re just vocalizing spontaneously. More description of physical interactions between the dolphins would allow a researcher to infer a more interactive sort of communication, but the author’s description of the dolphins just resting at the surface ‘as if dozing’ during the recording provides no information from which to make such an assumption. However, without any of this information, a reader of the 2016 publication has no choice but to take Ryabov’s misrepresentation of the calling behavior during the study as truth.
The second incorrect claim: the possible presence of discrete pulse types means it has to be language.
This part of the discussion is where both the 2011 and 2016 papers really start breaking down in terms of credibility. It’s important to note that this paper was published first in 2011 in an open journal of acoustics, which means that the bar for peer review and thorough scientific inquiry is pretty low. It was republished in 2106 in a journal of physics and mathematics, which is a very odd place to see an ethological/linguistic paper. Both of these places of publication call into question the validity of the work, as no author would choose to publish them in those journals if they could get into more well-reputed ethological or linguistic/acoustic ones.
The discussion section of the 2011 paper makes a very weird logical jump from the reasonable assumption that the dolphins are producing sounds to communicate with each other (although there’s a the possibility that these vocalizations are self-referential) to the calls being phonemes and being part of word production just like in human language. It’s a huge, illogical jump with very little evidence to support it - even if they seem analogous, that doesn’t mean it’s the same pattern. Let’s look at the direct quote from the 2011 paper:
“The non-coherent pulses produced by the dolphins differ from each other by the waveform in time domain and by the set of spectral components in frequency domain (Figure 3). In this connection it is possible to assume, that each non-coherent pulse represents the phoneme of the speech of a dolphin spoken language and every pack of the non-coherent pulses is a word, then sequence of the packs of the non-coherent pulses is the sentence. For the best understanding, we shall compare speech of a dolphin spoken language with speech of a human spoken language. (Ryabov, 2016)"
First off, there’s no real evidence within the 2011 paper or its citations that these discrete types of pulse patterns have been recorded elsewhere - no pulse pattern was repeated throughout the data in this study according to the 2016 paper - which means that the author is just assuming they’re stereotyped and regular enough to have singular meanings. Second, that quote contains a bunch of false equivalencies without any data to support them: that pulse calls have discrete meanings, that combinations of pulse calls has meaning, that the meaning of the combination of pulse calls is more complex than that of the individual units that comprise it, that a pack of pulses is yet another larger information unit with more inherent complexity, and that it is similar if not analogous to human language structures. Past the knowledge that apparently discrete clicks exist, there is no data to support any of those assumptions - yet they're the ones that the conclusions of both papers and the media furor are predicated upon. The 2016 paper is slightly more tactful about how it approaches comparing the assumed dolphin phonemes and words to human speech, stating that it ‘seems interesting to compare them’ rather than directly assuming human language is the best comparison. However, the discussion section still massively reinforces the idea that dolphins have been proven to produce words by treating that hypothesis as fact for the duration of the dolphin/human comparisons.
But wait, there's more:
From there, Ryabov makes all sorts of even more expansive claims about how the language structure he’s “discovered” in dolphins fits the traditional criteria for language as humans think of it, how features of language ‘directly correlate’ to intelligence and consciousness, and how other unrelated learning and communication studies with dolphins suddenly support what he’s discovered. He speculates wildly about how dolphins might form words using a combination of multiple sounds within a specific pulse, but has nothing to cite to support that in the slightest. All of these claims are listed as ‘assumptions’ about things he has no data to support, but are written about as if they are fact. Considering how all the assumptions those statements are predicated upon have already been explained as misleading if not outright wrong, there’s no point in going into further detail about Ryabov’s continued erroneous exposition.
Occam's Bikini and the media's fascination with humanizing non-human animals:
At the end of the day, none of Ryabov’s research really imparts any new information. Dolphins communicate with each other, which we knew already - all he actually managed to show concretely is that click-train calls come in coherent and noncoherent bursts. He’s made a reasonable hypothesis that these different types of bursts might impart different information to a listener, but that’s it. Literally everything else Ryabov's papers appears to be derived from assumptions, guesswork, and false equivalencies dressed up nicely in formal language. The 2016 paper appears to have been expressly re-written to take out the very important inconclusive data so it would make a splash with the media and bring in more funding. Guess what? It did. The news agencies took it and ran with it - and even the few like CNN that actually read the 2016 paper and interviewed sources who disagreed didn’t look critically enough at the original publication to catch the glaring inconsistencies and lack of scientific rigor. Now it’s all over the internet and, like so many incorrect myths about animal behavior, will probably never disappear. This is a perfect example of the Occams’ Bikini mentality- the sexiest solution, not the correct one, is the one the public will prefer.
Dolphins aren’t communicating in sentences. They’re talking over each other all the time. They’re certainly not being chivalrous about letting female dolphins repeat themselves in conversation. Dolphins are communicating like dolphins - not like humans - with whistles and clicks and screams, and exactly in what way they're transmitting information is still unknown.
Sources and further reading:
Communication - Dolphin Research Center. (n.d.). Retrieved September 14, 2016, from http://www.dolphins.org/communication
Gregg, J. (n.d.). How Do Dolphins Communicate? Retrieved September 14, 2016, from http://www.dolphincommunicationproject.org/index.php/2014-10-21-00-13-26/dolphin-communication
Trone, M., Balestriero, R., & Glotin, H. 6.1 Gabor Scalogram Reveals Formants in High-Frequency Dolphin Clicks.
Ryabov, V. (2011). Some aspects of analysis of dolphins’ acoustical signals.Open Journal of Acoustics, 1(02), 41.
Ryabov, V. A. (2016). The study of acoustic signals and the supposed spoken language of the dolphins. St. Petersburg Polytechnical University Journal: Physics and Mathematics.
Westcott, B. (2016, September 13). Dolphins may have a spoken language, new research suggests. CNN. Retrieved September 14, 2016, from http://www.cnn.com/2016/09/13/europe/dolphin-language-conversation-research/