Dogs Don’t Process Language With Their Left Brains, After All

By Neuroskeptic | April 7, 2017 12:53 pm

A case of left-right confusion misled researchers about how dogs process language.

dog_left_brainLast August, Hungarian neuroscientists Atilla Andics and colleagues reported that the left hemisphere of the dog brain is selectively activated in response to the lexical properties (i.e. the meaning) of spoken words. This result was very interesting, not least because lexical processing is also lateralized to the left hemisphere in most humans. The paper appeared in the prestigious journal Science.

However, in an “Erratum” published today, Andics et al. reveal that they had mixed up the left and right orientation in all of the dogs fMRI images. In other words, the “left hemisphere” activations were actually right hemisphere ones.

In fairness to the authors, keeping track of the orientation of MRI images is by no means trivial. It is impossible to tell, from looking at an MRI of a healthy brain, which side is which. So, if a left/right confusion occurs, no-one would notice it. The MRI scanner should log the left/right orientation of the images at the time of scanning, but this data is often stored separately from the images themselves, in a “header”, and this header can be lost or corrupted during image processing.

Andics et al. say that the confusion in this case happened “in the process of accounting for the different body positions of humans and dogs in the MRI scanner.” If I had to guess what happened, I’d say it could have been a case of “over-correction”. Maybe someone thought that the images were the wrong way around, and flipped them to ‘correct them’, when in fact the orientation had already been corrected by someone else. A case of the left hand not knowing what the right is doing, in other words.

Whatever happened, the error is one that any fMRI researcher will be able to empathize with, and the authors deserve a lot of credit for coming clean. I’m sure that a lot of researchers have discovered this kind of error in one of their old papers and just chosen to keep quiet about it.

On the other hand, I think Andics et al. are pushing it a bit when they say in their Erratum that “This error does not affect the main conclusions of the paper.” After all, the second paragraph of the original paper makes much of the fact that “Lexical processing in humans is lateralized to the left hemisphere”, and at the end of the original, the authors wrote that:

We discovered three neural mechanisms of speech processing in dogs. First, there was a LH [left hemisphere] bias for processing meaningful words, independently of intonation. Second, acoustic cues of affective speech intonation were processed independently of word meaning in R mESG [right middle ectosylvian gyrus], and intonational markedness increased functional connectivity between auditory and caudate regions. Third, dogs relied on both word meaning and intonation when processing the reward value of verbal utterances. All three findings reveal functional analogies between dog and human brain mechanisms.

I’d say that the supposed left lateralization of lexical processing in dogs was a ‘main conclusion’ of the original paper, or at any rate, pretty important to it.

Science have now updated the Andics et al. paper. The new version is the same as the original but with “left” replaced with “right” in various places. In other places the reference to left/right has been deleted, for instance the new abstract says “We found a hemispheric bias for processing meaningful words”, where the original said “We found a left-hemisphere bias for processing meaningful words.”

Indeed, from reading the revised paper you’d be forgiven for thinking that all of the results make sense and fit with the previous literature. But this was also the case with the original paper – a reminder that it’s possible to make sense of almost any results if you try hard enough.

CATEGORIZED UNDER: animals, fMRI, select, Top Posts
ADVERTISEMENT
  • https://about.me/ccnomad Catherine

    Do dogs have eq of Broca’s & Wernicke’s areas?

    • 9eyedeel

      yes indeed, but in the canine, they are referred to as the “Barker’s” and the “Wienerdog’s” areas, respectively

  • Pingback: Dogs Don't Process Language With Their Left Brains, After All - Artificial Intelligence Online()

  • basneggers

    In DICOM and the most common form of Nifti (.nii), full voxel to world information and other header info is included in the files itself and not in separate header files. This covers the vast majority of MRI data containers in use these days. So losing header files is not an excuse anymore since Analyze format got out of fashion, which was around 2005.
    Sloppy processing pipelines, preventable coding mistakes and the lack of a proper code testing strategy are the most common cause of L/R confusion these days. Always test your pipelines first with MRIs of a head with a vitamine capsule glued to the left or right forehead is quite common advice where im working. Never trust your tools blindly.
    So no, this is not an excusable mistake IMHO. To err is human, to forgive divine, of course, but this points out once more that your science is only as good as your instruments are.

    • practiCalfMRI

      Yes, it is very important to have some sort of fiducial marker like a vit. E capsule, used consistently. Registering a dog as “head first supine” (which is probably what they did) will ensure the header and orientation issue is consistently wrong, whether or not it’s with the image data. A fiducial inside the image is the only way to be 100% sure.

      • basneggers

        No, that is not entirely true. The header is always right, the DICOM tags ‘ImagePositionPatient’ (0020,0032) and ‘ImageOrientationPatient’ (0020,0037) uniquely define the relationship between each voxel and the mm x,y,z position axis system fixed with respect to the scanner bore, in a LPS gimbal. When you do not realize how your specimen was placed in the bore, that is entirely your fault.

        See https://www.slicer.org/wiki/Coordinate_systems for a good explanation on spacing of MRI images, and the DICOM tags in the specs of the standard itself: ftp://dicom.nema.org/MEDICAL/dicom/2015b/output/chtml/part03/sect_C.7.6.2.html

        Something similar holds for Nifti, in a way (full affine voxel2world matrix is the header fields). See https://nifti.nimh.nih.gov/nifti-1 for more information.

        It really is the user who gets the images wrong, or a badly designed viewer. Whether a specimen is upside down in the scanner or not.

        My point is, the data format is not to blame. It is the software and/or the user that interprets it that *could* do things wrong (often written by researchers). In principle the fiducial should not be needed when space is dealt with consistently by the processing/viewing software.

        But I agree that it is good advice to place fiducials, as we do not live in software utopia yet. We always do it for at least a few inclusions in a study, to verify.

        • practiCalfMRI

          “When you do not realize how your specimen was placed in the bore, that is entirely your fault.”

          “My point is, the data format is not to blame.”

          This is what I meant. I should have said “consistently incorrect” rather than “consistently wrong.” It’s not the header’s fault, as you say, but by setting the reference frame incorrectly the header now contains incorrect information.

          • http://secure93.com Joel King

            yes I pulled successfulness to take home Ninty Eight Thousand Five Hundred and Sixty Seven dollars.I heard of this job 6 months past and a pair of months prior I already have tested and I have obtained such a great wealth. It is very good for myself and my family.My siblings wondered to see how exactly I settled my online career and my Life in few time. Thanks a lot to GOD for all this….You also can obtain similar number of funds, if you have a PC and web connection at your home. different complicated skills are not required to undertake this work. I am Loving it and all of you will also like it and you will definitely not forgive yourself if you not check out the information >>>>>> B1Z.ORG/37Y

  • Lord

    Maybe because people lay on their backs and dogs on their stomachs?

    • William Herrera

      Yes, and perhaps the readers of the scans were so unacquainted with canine anatomy they failed to realize the heads were upside down relative to the human heads in such a scanner.

    • practiCalfMRI

      Head first supine is incorrect for dogs in the sphinx position. Oops.

  • Leonid Schneider

    I am confused. Why do dogs need to understand the meaning of human language? Yes, they are the only animals which care about our feelings and expectations, but dogs are unlikely to do so by learning English, Russian or Hungarian from their owner. They do read our facial expressions, eye and body movements, tone of the voice and bodily transpirations much better than other humans do though.

    • jrkrideau

      Dogs ‘understand’ human vocal commands so they must need to process speech to some extent. As an experiment, if your dog was trained using English try some German or Russian language commands.

      • Leonid Schneider

        I never said dogs were deaf. Associating a certain sound sequence to something is not the same as processing speech, a method of communication which only humans have as such.

  • smut clyde

    If dog-brain functions were lateralised with some kind of species-wide hemispheric dominance, you would think it would show up in other ways that didn’t require a fMRI scan to detect.

  • floatingnotes

    Wasn’t there another left/right correction in a prominent journal within the last few years ?

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Hmm, not that I can remember, but you might be right

  • storkchen

    Either way, this is bad science.

  • Pingback: Dogs Don't Process Language With Their Left Brains, After All - Neuroskeptic()

  • Pingback: Cross-blogging del sabato - Ocasapiens - Blog - Repubblica.it()

  • practiCalfMRI

    Same problem in a 2014 Current Biology paper from the group: http://www.sciencedirect.com/science/article/pii/S0960982217303342

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Presumably these are all consequences of one single mistake that happened years ago. This is an example of what I call the Tufnel Effect: “It’s such a fine line between stupid and…uh, clever.”

  • Attila Andics

    Thank you for your article on our Erratum. Yes, the error was an unfortunate “over-correction” basically. The left-right flip we applied to dog images to account for their orientation did not change the images’ relation to the coordinate system because, erroneously, that was left-right flipped as well. We apologize for this.

    We think that the main conclusions of the paper are indeed not affected. The important aspect about dogs’ hemispheric bias is not its direction but its lexical nature. Lateralization for processing sounds with communicative value has been found in several mammals, and the direction of this asymmetry appears to vary across species (cf. Gil-da-Costa and Hauser, 2006). We found right hemispheric bias in dogs for processing meaningful but not meaningless words, independently of intonation. Regardless of direction, this is evidence for lexical processing and reveals a functional analogy between dogs and humans. Other conclusions involved no claim of significant lateralization.

    So the main conclusions are not affected, but I think the corrected version is clearer about where the real emphases are.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Thanks for your comment – and again, well done for making the decision to correct this error.

  • CL

    Thanks to the scientists for spotting and correcting their error. I know of several papers and research groups that have not corrected know left-right errors. I suggest using a fiducial all the time.

  • Pingback: TCTS weekly: от летних школ по нейронауке до таинственных препринтов | THINK COGNITIVE, THINK SCIENCE()

  • Pavel Prosselkov

    This is not just a simple left-right sides switching, it makes a completely different story, which is no longer as valid as Science sells it. Anatomical asymmetry underlies the functionality of the signal processing bias but there is no way it explains it

NEW ON DISCOVER
OPEN
CITIZEN SCIENCE
ADVERTISEMENT

Neuroskeptic

No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.

ADVERTISEMENT

See More

@Neuro_Skeptic on Twitter

ADVERTISEMENT

Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar
+