What's the use of Grammar?

I'd like to offer an alternative perspective to all the answers offered so far. First, I'd like to rephrase the question and then react to the other answers with some general points.

1. When you ask what is the purpose of grammar, you really want to know what is the purpose of morphology? And, in fact, this is one of linguistics' great puzzles (although not one that is often discussed openly). By and large morphology is redundant in all languages. As you say, it does not matter whether you say 'John likes' or 'John like'. In fact there are many dialects of English where 'John like' is just fine. There's certainly no potential for misunderstanding. The vast majority of inflectional morphemes are not necessary to convey the desired meaning. That does not mean that they do not convey meaning that would not be lost without them (e.g. in Czech accusative and locative often differentiate between a directional and locational meaning of a preposition or even the English plural morphemes are pretty useful) but that the meaning is generally sufficiently clear from context or could be easily recovered through some other means. There are various very powerful models that explain the process of the emergence of inflectional morphology but no clear explanation as to why it's there.

2. So by and large you can do without morphology but can you do without syntax? That depends on what you think syntax is. If you think about it as mostly word order, then you always need some way to order words one after another. And there does not seem to be one natural, logical way to do it. Subject-Verb-Object may seem natural to us but any other ordering seems to be just as fine. But how necessary is it? Not very. Most sentences are fine in pretty much any order. There are some high profile examples where the sentence alone is completely ambiguous (e.g. Parents love children.) but they would generally be rescued by context (e.g. in Czech 'Parents love children' is always ambiguous without context and in English, you're more likely to disambiguate by including 'their' before the object). But even in languages with relatively fixed word orders, there are many contexts (like poetry) when it can all be upended and mixed beyond recognition.

3. While there are no languages without some sort of word order conventions, surely syntax is not about the order of words, at all! It's all about capturing the underlying relationships between words and giving them some sort of communicative or configurational function. Nouns become subjects to express agentivity, adjectives become attributes, verbs predicates, etc. We need to capture how agents and patients relate to each other, make sure we know what it is that prepositions modify, link anaphors to their referents, etc. And sure enough, all languages do that in one way or another. But they all do it in a profoundly redundant manner. Almost anything that you might think is essential for communication based on your observations of one language will be found missing in another language. All attempts at finding grammatical universals have failed to find anything plausibly invariable. Even the famous 'Universal Grammar' is just a collection of a handful of some general tendencies.

4. All the answers to the question, have assumed that there is such a thing as grammar. And not just that. They also assume that grammar is the fundamental expression or carrier of the deeper purposes of language. But I'd like to propose that there's in fact no such thing as grammar. Grammar is consequence of its artifacts, the grammar books. I've written a long post about thinking about grammar in different ways.

5. Construction grammar offers a much better way of thinking about grammar as an inventory of conventional ways of expressing meanings through sounds, words, parts of words, and combinations of words (constructions). And it's relatively easy to explain why we want to have conventional ways of speaking. Everything in human interaction is governed by convention (tacit agreement). In makes the interaction more predictable, easier to teach to new generations, more useful in differentiating in-groups from out-groups, and of course, it also make interaction more efficient. Not having to negotiate your conventions at the start of every interaction is very helpful. Even redundancy makes great sense because you often deal with the transmission of information over noisy channels (e.g. in a noisy place). Of course, over time, conventions tend to pile on top of one another and the efficiency will suffer. But no matter how morphologically or syntactically complex, no language seems to have transcended the equilibrium of efficiency.

This last point has to deal with two important objections:

i. What about the obviously different cognitive natures of the conventions of grammar when compared to the conventions of culture? It's much more difficult (and quite different) to 'learn a language' than it is to 'learn a culture'. Our ability to behave in conventional linguistic ways is embedded so much 'deeper' in us than our ability to behave in culturally conventional ways. Just compare acquiring the phonological distinctions of a language with learning to politely refuse an offer of food. This is a good point to which I only have a partially satisfactory answer in that the conventions of culture are much deeper and more complex than people imagine. But I have to admit that they still seem qualitatively different from the conventions of language. I've addressed some of this in another post.

ii. Isn't talking about an inventory of conventional ways of expression just renaming what we've always talked about as grammar? Yes and no. In claiming that grammar is just an artifact of the grammar book, I'm not suggesting that the grammar book is a deluded enterprise. For millenia, grammar book writers have been describing real conventions and have done a great job in putting together taxonomies of these conventions. So when rethinking the nature of language without 'grammar', we don't need to start over and jettison all those old rules. They're pretty handy for what they are. But we need to rethink their place in our understanding of language. We must not confuse all those taxonomies of linguistic conventions with language for which we then need to find cognitive and neurological correlates. These taxonomies that were developed for the convenience of the reader and writer (and often influenced as much by the convention of the genre as anything else). The structure of language does NOT look like the table of contents of a grammar book. There's probably not even such a thing as a single structure. But that is often what we're lead to believe.

So in conclusion, even though this question may seem a bit naive, it is actually a profound puzzle to which linguistics does not have a complete answer. The fact, that we can be so dismissive of the question only indicates how much we need to rethink what we've been taking as givens about language for at least a hundred years. There are many linguists and other cognitive scientists who are doing that. But perhaps few who would want to be as radical about it as this. But I suspect that in a few generations, this will not seem as bold, as it may now.