The dental fricatives (/θ/ and /ð/; spelled with th) often present a challenge to non-native learners of English. Depending on the speaker's native language, different phonemes may be substituted. In Indian accented English, the dental fricatives are replaced with some variation of /t/ or /tʰ/; native Mandarin speakers substitute /s/ or /z/. (source)

This is surprising because both Hindi and Mandarin have consonants close to /s/, /z/, /t/, and /tʰ/. What causes Hindi and Mandarin speakers to replace /θ/ with different sounds then?

  • My hunch is that, although /t/ is described as laminal denti-alveolar in both Hindi and Chinese, Hindi also has retroflex /ʈ/, which may be reinforcing the place of /t/ as a dental consonant, which makes /t/ even more of a suitable substitution for /θ/.
    – Nardog
    Commented Aug 9, 2018 at 0:19

4 Answers 4


Here's a paper that's addressed a similar phenomenon of the different realizations of /θ/ between Cantonese and Sichuanese speakers, both of which are dialects of Chinese and share similar phonetic inventories.

The paper conducted production and perception comparisons between Cantonese and Sichuanese native speakers as to explore the reason for different choices of realization for /θ/ in English as L2.

The production tests came out somewhat insignificant. However, in perception tests, Cantonese speakers showed a significantly lower accuracy rates of discrimination between /θ/ and /f/ than Sichuanese speakers, and Sichuanese speakers between /θ/ and /s/, which aligned with their respective choice of realizations as discovered in a pilot test.

And the author argued with comparisons of the function load of different phonemes in two dialects that different outputs for one L2 from different L1 speakers are rather frequency-motivated than markedness-motivated.

  • Interesting paper! I don't quite understand the section on functional load -- they showed that /s vs f/ distinction was more important in Sichuanese and less important in Cantonese, but why does it follow that Cantonese speakers prefer /θ/ > /f/ and Sichuanese speakers prefer /θ/ > /s/?
    – Bai Li
    Commented Aug 7, 2018 at 4:10
  • @Bai My understanding is as the lower function load of /s vs θ/ in English compared with /s vs f/ yields more confusion between /s/ and /θ/ instead of /s/ and /f/, the same would apply if we could do the same comparison in Chinese. But given the absence of /θ/ in Chinese, the lower function load of /s vs f/ in Cantonese gives a higher inclination for Cantonese speakers to override L1 preference than Sichuanese speakers, since we cannot compare the function load for any pairs including /θ/ in Chinese.
    – Matthew Su
    Commented Aug 7, 2018 at 6:42
  • And I'd like to add that I'm not comfortable with the phrase 'functionally more loaded' for a single phoneme as used in the aforementioned paper, where the author tried to link how much one phoneme is fuctionally loaded to its preference in realization. Firstly because it's seldom characterized that way and secondly function load is a parameter for a opposition pair, never one phoneme.
    – Matthew Su
    Commented Aug 7, 2018 at 6:48
  • Still, differences in L1 markedness (and frequency is still a markedness) are a very satisfying explanation for the phenomenon. Commented Aug 7, 2018 at 9:05

This is a great question without a clear answer. People have struggled to find the answer since the 1970s:

Here is my 2002 paper with many references listed in Appendix A. See also my dissertation on the topic.

There are many cases of two languages or two dialects with the same or similar phonological inventories (the same phonemes), which substitute different sounds in place of /θ/, usually either [t, f, s]. The consonant [t] appears to be the most frequent substitute, and this is usually attributed to it being the "unmarked" consonant (most frequent cross-linguistically, easy to articulate). Researchers have proposed many different solutions, but as far as I know, there is little consensus. Some have proposed that different dialects have different grammars (e.g. Weinberger 1988, Brown 1998, Lombardi 2003). Others say that it depends on "functional load" (e.g. Hancin-Bhatt 1994): The functional load of a given feature value (e.g. +continuant) is equivalent to the number of times it is specified in the phonemic inventory divided by the total number of phonemes. Note that this approach does not account for different substitutions in two dialects with the same phonemic inventory (e.g. France French [s] vs. Québec French [t]).

Still others say that it is not the phonemic inventory that is important, but instead it is the phonetic implementation that matters (e.g. Brannen 2002, 2011). For example, the actual pronunciation of /t/ can be different from one dialect to another, e.g. English has an alveolar /t/, but many languages have a dental /t/ (closer to the place of articulation of [θ] which would make dental [t] a good substitute.)

One thing that is clear from the research however is that this is not simply a production phenomenon: speakers of languages that do not have [θ] misperceive this sound differently, and it depends on their native language.

  • Brown, C. (1998). The role of the L1 grammar in the L2 acquisition of segmental structure. Second Language Research , 14, 136-193.

  • Hancin-Bhatt, B. (1994b). Segment transfer: A consequence of a dynamic system. Second Language Research , 10 (3), 241-269.

  • Lombardi, L. (2003). Second language data and constraints on Manner: Explaining substitutions for the English. Second Language Research , 19 (3), 225-250.

  • Weinberger, S. H. (1988). Theoretical Foundations of Second Language Phonology. PhD Dissertation . University of Washington.


Quoting the paper you link:

A particular sound which does not exist in the native language can therefore pose a difficulty for the second language learners to produce or some times to try to substitute those sounds with similar ones in their mother tongue. These sounds include both vowels and consonants. For example, there are no vowels like /æ/, /au/, and /ɛə/, etc. or no such consonants as /ð/, /θ/. Therefore learners have trouble first of all in perceiving these sounds, and consequently try to find nearest equivalents to substitute those new sounds. A typical example will be the substitution of /s/ or /z/ for the English /ð/, /ai/ or /e/ for the English /æ/ as in the word ‘that’.

The answer is simply that in each language, these sounds are perceived to be closer to the English sounds than other sounds available in their inventory:

  • /θ/ /ð/ are dental fricatives
  • [s] [z] (the common mandarin realisations) are alveolar fricatives, so that which is different is the place of articulation
  • [t̪ʰ] [d̪] (the common hindi realisations) are dental plosives, so that which is different is manner of articulation

enter image description here

  • 7
    Yes, but the question is why exactly Indian speakers pick /t/ while Mandarin speakers pick /s/ (assuming that's actually the case), even though both languages have both choices available. Why only Mandarin speakers feel /s/ to be closer? Why isn't it the other way around? Is there anything about the realization of Mandarin /t/, /s/ and Hindi /t/, /s/ that makes Mandarin speakers, when faced with the choice, pick /s/ as the closest to /θ/, while Hindi speakers choose /t/ instead? Or is it some historical reason, or pure chance? Commented Aug 5, 2018 at 12:06
  • 4
    To some degree it's convention. English and German names starting with 'H' used to be rendered in Russian with /g/ (consider the name 'Goffman') but are now usually rendered with /x/.
    – Colin Fine
    Commented Aug 5, 2018 at 18:39

I think this is caused by the existing transliteration rules for writing Hindi words in Roman letters. The Hindi letter थ (/t̪ʰ/) is written as th when writing Hindi words in Roman letters. For example: थोडा is written as thoda which means 'less', and a name आदिनाथ is written as Adinath. But स (/s/) transliterates to s. Eg: सब - sab which means 'all'. Also, ज़ (/z/) transliterates to z. Eg: ज़माना - zamana which means present generation.

It thus makes sense that new learners who have not heard the pronunciation /θ/ will associate th (in through, for example) with the available /t̪ʰ/ sound rather that /s/ or /z/ sounds. This is especially true in the days the populace had limited access to TV shows from US or UK to observe their accent.

However, the th in through is a voiceless sound, but the th in that is markedly a voiced consonant. The latter is therefore pronounced with an available voiced sound /d̪/ (of the Hindi letter द) which falls in the dental consonant cluster of the Hindi alphabet, just like /t̪ʰ/.

Transliteration is quite an important factor in the Indian context, since most of India uses English in the written form in official work, and names are transliterated in English for ID cards, etc.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.