Consider the following sentence:

"This paper introduces a new alternative for generating synthetic data based on images."

What I want to say is that "the new alternative" is "based on images". Thus, the noveltiy of the new alternative is in the fact that it is based on images. What I would like to avoid is that one think that I mean "image-based synthetic data generation already exists and I propose a new alternative to it."

So, to avoid the ambiguity I add ", which" to the above sentence, so it becomes:

"This paper introduces a new alternative for generating synthetic data, which is based on images."

However, after looking for the usage of ", which", I am still not sure that the sentence means what I mean. More precsisely does "which is based on images" refer to "generating synthetic data" or to "new alternative"?

2 Answers 2


I’d say that most readers would interpret it in the way you intend.

If you are highly concerned to be understood correctly, you could always recast the sentence along then lines of “This paper introduces a new approach for generating synthetic data that—unlike existing approaches—is based on images.”


Your alternative, with the comma

  • This paper introduces a new alternative for generating synthetic data, which is based on images.

uses the non-defining relative convention

  • John is my pen-friend, who lives in Canada. = John is my pen-friend; he lives in Canada.

as opposed to the defining convention

  • John is my pen-friend who lives in Canada. = John is the pen-friend of mine who lives in Canada (not the ones in France, Tibet ...).

However, there is intrinsically still the ambiguity over whether this means that 'a new alternative for generating synthetic data' or 'synthetic data' itself is based on images. While this ambiguity is doubtless obviously resolved by logic in this case by most educated readers, the ambiguity 'which preceding noun phrase does the relative clause refer to?' is quite common and sometimes confusing.


Your relative clause if written without the comma would be defining, but must now really mean 'for generating image-based synthetic data'.


The original 'reduced relative clause' (some analysts disagree with the name/analysis) is interestingly ambiguous, either defining referring to 'synthetic data' (' ... image-based synthetic data') or non-defining referring to 'a new alternative for generating synthetic data'. The comma convention weakens with trickier examples. Reading these out, stress patterns would disambiguate: 'synthetic data based on images' read without a pause or stress on 'based on images' forces the defining sense (= ' ... image-based synthetic data'), with 'synthetic data' the modified NP. Stressing 'based on images' after a pause switches the meaning to [this new alternative is, incidentally, based on images']. But speech is more flexible than printed material.

Your Answer

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

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