Hearing bad grammar affects your heart

Linguistic knowledge and skills develop largely without awareness and may therefore be difficult or impossible to articulate. A recent paper examined whether variability in heart rate can be used to assess implicit language knowledge.

The paper is Physiological responses and cognitive behaviours: Measures of heart rate variability index language knowledge, Dagmar Divjak, Hui Sun, Petar Milin, Journal of Neurolinguistics, vol. 69, 101177. https://doi.org/10.1016/j.jneuroling.2023.101177


41 healthy, adult speakers of British English listened to 40 short English speech samples. Half of the samples contained grammatical errors, affecting between 18% and 56% of the nouns in the sample. The errors related only to use of definite or indefinite articles (the, a, an).

The research looked at English article errors because these errors:

  • are frequent in the English speech of non-native speakers.
  • are not usually corrected, because they tend not to impede communication.

The authors created each speech sample from a transcribed response to an interview question, extracted from the BACKBONE4 English as Lingua Franca (ELF) Corpus of Polish speakers. In the samples containing errors, the authors inserted additional English article errors to bring the error density to the desired levels They removed errors not relating to articles.

Polish is a language with no articles. It is not surprising if Polish speakers find it difficult to master the use of articles in English.

Example of some errors

Example A shows one sample containing errors (with article errors and the related nouns underlined).

Example A. with errors
I think that culture is one of the areas most affected by a globalisation and it’s hard to say whether it is the positive or negative impact. I think that thanks to a globalisation, people all around the world listen to same music, watch the same movies, and read same books. They can discuss the same issues with each other, and understand each other better, because they know what they are talking about.

As example A shows:

  • some articles were superfluous (a globalisation).
  • some articles were missing (   same music/books).
  • some definite articles replace indefinite articles, or vice versa (the positive or negative impact).


High heart rate variability reveals a highly irregular heart rate.  The researchers found a statistically significant reduction in HRV in response to samples containing errors. The reduction was proportional to the number of errors, up to the point where the error rate reached about 40% of the nouns heard. Beyond that point, heart rate variability was roughly flat.

Participants were also asked to describe in their own words the type of errors they heard:

  • one participant reported that there were no errors.
  • no single participant identified errors in article as the only error.
  • only 7 participants did not mention article errors at all. But many participants were unaware of the name of the category, stating instead that “the letter ‘a’ was missing” or that ‘the’ was “forgotten or added in unnecessary places”.
  • only 2 participants selected/suggested ‘incorrect use of articles’ as the only category heard (from a list of 6 errors) alongside ‘incorrect pronunciation’. And 4 participants did not select ‘incorrect use of articles’ at all.

As the researchers had expected, these results confirm that participants have mainly only implicit linguistic knowledge and are not able to express that knowledge explicitly.

Each participant heard speech samples read by four speakers: male and female native speakers of British English and male and female foreign speakers (Polish). Participants estimated the number of errors as much lower for the British female voice (mean = 16.5, standard deviation [SD] = 18.5) than for the others: British male speaker (mean = 23.7, SD = 25.4); Polish female (mean = 22.20, SD = 18.5); Polish male speaker (mean = 22.83, SD = 18.67).

After participants completed the task, they were asked about the origins of the foreign accent; 5 did not identify the accent as Polish.

Sensitivity to accent

In addition to sensitivity to error density, participants were also sensitive to accent. Their heart rate variability dropped more on hearing errors by British speakers than on hearing errors by Polish speakers. The researchers say that this is because errors are more unexpected in native-accented speech than in foreign-accented speech.


Theoretical linguists often rely on speaker’s metalinguistic linguistic ability to, for example, judge whether a particular sentence is grammatical. However, language users typically:

  • produce and process language in real time for communicative purposes, without trying to assess whether sentences are grammatical. They have only limited metalinguistic awareness—ability to focus attention on language as an object in and of itself, to reflect upon language, and to evaluate it.
  • acquired their first language unconsciously. As a result, their knowledge of their first language is largely implicit and they deploy that knowledge without conscious awareness.

The authors say their paper provides the first evidence that heart rate variability can reveal a person’s implicit linguistic knowledge without the person needing to articulate that knowledge explicitly. They suggest that measures of HRV measures can:

  • reveal linguistic knowledge that language users hold implicitly, without unjustifiably relying on language users having metalinguistic ability to articulate their knowledge explicitly.
  • be made using portable and non-intrusive systems usable in natural environments and with a wide range of populations.
  • avoid bias caused when language users try to make explicit metalinguistic judgments and those judgments are influenced by education about which forms are and are not considered acceptable.


  1. I must be missing something. I would have expected a statistically significant INCREASE in HRV in response to sample containing errors.

    “The researchers found a statistically significant reduction in HRV in response to samples containing errors. The reduction was proportional to the number of errors, up to the point where the error rate reached about 40% of the nouns heard.”

    1. I had the same reaction. I’ve checked the paper again and I have summarised the direction correctly.
      The Discussion section (4) includes the following:
      ‘Expectation violations tend to require more cognitive effort, thereby putting stress on the cognitive system, and this is picked up by the physiological system. Recall that the applied literature generally accepts that HRV indexes stress: mental stress leads to an increase in interval regularity and thus a decrease in heart rate variability ‘
      There is also some discussion in section 1.2, but I still don’t understand why it is that way round.

Leave a comment

Your email address will not be published. Required fields are marked *