Isn't that an excellent question? I have heard that a lot. A lot of people have told me Chinese has no grammar. I am like 'huh? Yeah, it kind of extremely does.'
Ah, maybe that's it...one of the students in my exchange program to China kept saying this, so finally I said but you are taking a Chinese class how can you think this
, and it turned out by "grammar" he meant "inflection."
Huh. So in the popular mind, inflection = grammar. Ew.
From what I understood from the people who told me this (mind you, they were not studying Chinese) is that apparently Chinese just throws words out there and somehow it all miraculously makes sense.
I think some people who do know that grammar is more than inflections - that grammar does include, for instance, word order and such - fail to understand that all languages do have grammar.
First of all, of course, and this is a generalization of the "inflection = grammar" problem, is that they fail to appreciate just how many ways grammar can exist - inflections, word order, lexical hierarchies
Re: lexical hierarchies, a good example are languages where in a clause with two nouns and one verb, it is very likely in that nouns of higher animacy are subjects, and nouns of lower animacy are objects; even in languages of that kind, though, there may be contextual cues as to when that assumption is disregarded as well even without explicit marking. Now, all of that *is* grammar, but try explaining that to someone who likes thinking that, say, Africans or Aboriginals or Asians don't have grammar - first of all, you're off on so abstract and untangible things - hierarchies of nouns! - that have very little to do with what people associate with whatever grammar they've been exposed to in school. Such hierarchies of nouns aren't exclusive to isolating languages in far-off lands, though, you can find similar things acting in how Swedish speakers determine whether the order of a sentence has been inverted for emphasis. (There, definiteness also plays a role, not just animacy. Is my intuition, at least - so don't trust my claim entirely on that either.)
People don't generally understand that congruence also can exist for other things than gender, definiteness and case either - so the Chinese classifier system, which is a reasonably complex system with congruence and a bit of other things thrown in, probably won't register as grammar to lots of people, it's just words that tend to occur with words, a bit like how you'd say "on the roof" and not than "in the roof" when talking of someone who is standing on top of your house, isn't it?
Turns out even that is grammar, of course, but since these are things people don't get wrong (and some people of course do get annoyed if you get some preposition wrong, even in cases when there's ample precedent for both prepositions being used in that construction - heck, some people will try correcting you even if you use a construction correctly because it looks close enough to a construction they know people sometimes get wrong... - and so on) that often, it's not taught as grammar the same way that, say, Latin conjugation is.
People have a very vague notion of what grammar is. But they don't really have any use for a correct notion either, so it's not like we really can do anything about it.
In a way this also has to do with the notion of how complex different languages are. The problem really is there's no meaningful metric that will give a reasonable objective reading - how do you measure the intricateness of the noun class hierarchies involved in a language, and what if there are complications like, say, some noun being considered way lower in one number than in another? How does a complex noun inflection compare to changes that pop up in words in certain positions in sentences (e.g. the semitic pausal forms), how does having an almost completely unpredictable derivation used to derive the perfect aspect form from the imperfect (e.g. Russian and some other slavic languages) compare to a rather complex ruleset for which participants in a sentence can be topicalized and which cannot. It's impossible to come up with a metric that takes all of it into context. And hence, it's bullshit to even talk of it.