Agree to disagree then. Simply because single words in a noisy place is not realistic. We’re not testing what we actually hear in a vacuum. That’s what the original Word Recognition Test in a quiet soundproof booth is for, and of course it has its usefulness and that’s why it’s done.
But what we’re talking about here is in the context of the signal to noise separation performance in a real world test with surrounding noise. We don’t need to do another Word Recognition Test in noise. It’s useless to me because in this context, it’s not about what you actually “hear” (like you said), it’s about what you “understand”. I actually don’t really care what I “hear”, as long as I “understand” what is being said to me in a noisy place. If it involves my brain filling in the missing words, that’s fine with me.
It’s because your brain uses MANY things to isolate and understand speech, not just one thing. It’s not just what it hears, but how it hears it, like the tonal distinction between the speaker’s voice (like low for male or high for female), the loudness (or softness) of the speaker’s voice compared to others, and maybe the frequency of the speaker’s speech (how fast or slow they speak), and even what the babbles sound like differently from the speaker’s speech. All of that helps the brain isolate and formulate and arrive at an understanding.
In fact, if you simply do single word test in noise like you propose, chances are the result you achieve may fare worse than the result of a real-life complete sentences testing. That’s because the adjustment to improve what you “hear” can be overly corrected more than necessary for you to “understand”, possibly toward a more detrimental result instead of a more helpful result.