Every claim in the Short, and the sentence behind it
C1: The team recorded more than 162,000 calls
The team recorded 22 bats continuously for 75 days, collecting 162,376 vocalizations.
75 days of continuous recordings of 22 bats (12 adults and 10 pups) yielded a dataset of 162,376 vocalizations
Source: Results.
C2: The analysis focused on calls from seven females
The team used 14,863 calls from seven adult females in its analysis.
We included in the analysis 14,863 vocalizations of 7 adult females
Source: Results.
C3: Roost calls often happen during squabbles
The paper describes nearly all Egyptian fruit bat communication calls in the roost as happening during aggressive one-on-one interactions, including disputes over food or perching and protests against mating attempts.
nearly all of the communication calls of the Egyptian fruit bat in the roost are emitted during aggressive pairwise interactions, involving squabbling over food or perching locations and protesting against mating attempts
Source: Introduction.
C4: The model identified callers above chance
The model identified the caller with 71% balanced accuracy, compared with a 14% chance level.
The emitters of the vocalizations were clearly identified with a balanced-accuracy (BA) of 71% (where chance level was 14%, i.e. one out of seven
Source: Results.
C5: The model classified the interaction context above chance
The model classified what was happening with 61% balanced accuracy, compared with 25% chance. The contexts included food, mating, perching or sleeping.
We could classify the behavioral context with a significant BA of 61% (chance was 25%
Source: Results.
C6: The calls held clues about the addressee
The model classified the addressee's sex with 64% balanced accuracy, compared with 50% chance.
the addressee’s sex could be classified with a BA of 64% (chance was 50%
Source: Results.
C7: The model predicted how the quarrel ended
From the call, the model predicted whether the pair split up or stayed together with 62% balanced accuracy, compared with 50% chance.
Surprisingly, we could predict the outcome of a quarrel above chance level, with BA of 62% (chance is 50%
Source: Results.
C8: The calls sound similar to human ears
The authors say the different calls sound very similar to people.
which sound very similar to the human ear
Source: Discussion.
C9: What bats do with this information is unknown
The study does not establish whether bats themselves distinguish or use the patterns the computer classified. The authors say the information's natural function remains unknown.
While the natural function of this information is yet to be discovered
Source: Discussion.
C10: Almost 90% of the vocalizations were excluded
The authors say their conservative approach was the main reason almost 90% of vocalizations were excluded from analysis because callers could not be confidently identified.
This conservative approach is the main cause for the exclusion of almost 90% of the vocalizations from our analysis.
Source: Materials and Methods.
What the study doesn't say
These percentages describe how the computer model classified the calls, compared with chance. They do not show that bats themselves can tell apart the categories the computer picked out. The authors say the natural function of this information remains unknown.
The analysis used calls from seven adult females. Context, addressee and outcome were tested only on those calls. A separate caller test combined recordings from this study and an earlier one, using 15 bats; it reached 56% balanced accuracy against a 7% chance level (Results).
The bats were recorded in captivity, not in a wild roost. The conservative caller-identification approach also excluded almost 90% of vocalizations from analysis.
The source
Prat, Y., Taub, M., & Yovel, Y. (2016). "Everyday bat vocalizations contain information about emitter, addressee, context, and behavior." Scientific Reports, 6, 39419.
DOI: 10.1038/srep39419. Full text.
Open access under CC BY 4.0.
The Short
Watch it on YouTube: https://youtu.be/SXCvAG0R1Ko