Trying to compose an answer to the essay's title question exposes two areas of ambiguity. On the one hand, there is considerable debate in the literature over just how much of our proficiency in using language involves actual conscious knowledge of something versus simply implicit unconscious capability. And on the other hand, there is considerable debate in the literature over just how "innate" is to be interpreted. As a result, a spectrum of answers can be found in the literature, going all the way from "None!" to "All!".
At the "None!" end of the spectrum can be found the answers provided by the "Standard Social Science Model". According to this alleged paradigm, the mind is a general-purpose cognitive device shaped almost entirely by culture. In this paradigm, the human mind is a blank slate programmed by culture, and biology is relatively unimportant in understanding human behavior. So aside from the innate nature of the brain as a general purpose computing device, there is nothing specifically innate to our knowledge of language. According to Pinker, there are a number of scientists and philosophers who support this paradigm. Including, notably, B.F.Skinner. Pinker argues persuasively that the Behaviorism of Skinner is based on this paradigm. According to the "None!" paradigm, language is learned behavior, with the conventions of semantics and syntax being acquired on the basis of standard modes of deductive, inductive, abductive, and statistical (pattern identification) reasoning from the evidence available within the learning environment.
At the other end of the spectrum we can find the answers provided by Pinker's "Language Instinct" and Chomsky's "Language Organ". Chomsky famously argued against Skinner, that there was simply not enough information available in the learning environment to support the premise that the acquisition of language was learned. The "poverty of stimulus" argument was used to promote the idea that our brains come pre-equipped by nature and evolution with specific language rules (called "Universal Grammer"). Over the years since Chomsky introduced this notion in 1959, the "All!" paradigm has been modified by several contributors to deal with various objections that have been raised. The most popular current model involves the notion that the "language organ" (or the "Universal Grammer") has evolved a set of "parameter switches" that learning can flip either way -- thus explaining the wide variation in actual languages.
The debate between the two ends of the spectrum has generated a wide range of empirical investigations into the nature of actual language syntax, and the manner of actual childhood language learning. Almost all of this empirical evidence supports the idea that there is definitely something innate in how humans master language - so the "None!" paradigm at the far end of the spectrum is almost certainly wrong. But the evidence accumulated is not entirely persuasive that the "All!' end of the spectrum is the only answer. The evolutionary story for a particular "language organ" is not persuasive. Michael Tomasello in his 1995 book review of Pinker's Language Instinct, and Scholz and Pullum in their 2006 article in Stainton's Contemporary Debates in Cognitive Science lay out the evidence offered for the innateness of language, and offer some rebuttal of the full blown "All!" position.
As a result there has been a growth of intermediate positions. These do not deny that there is something innate beneath our mastery of language. They merely do not accept Chomsky's suggestion of a dedicated language "organ" as the necessary innateness. Here is where the ambiguity of the word "innate" surfaces. If the word is read to apply only to language-specific genetic endowment of an "organ" (of some sort), then theories in the middle of the spectrum would not count as thinking that language knowledge is innate. But if the word is read to include any capability of the brain provided by genetic endowment, then because they make use of more general-use capabilities of the brain, these theories would count as thinking that language knowledge is innate.
A lot of impetus for these intermediate approaches has come from investigations into the pattern recognition and statistical analysis capability of a "massively parallel connectionist architecture". And from the neurobiological study of the actual connections and modularity of the brain. If the brain does not have a specifically evolved "language organ", then it may achieve it's remarkable abilities at acquiring language by making use of a set of more general purpose "thinking" devices. There is no end of alternative theories along the spectrum. Where they might fall along that spectrum will be determined by just how "language specific" the theory posits the innate architectures to be.
Once one gives up a principled commitment to either end of this spectrum (both the "None!" and the "All!" answers), the resolution of the essay's title question becomes a matter of empirical investigation. The debate has ceased to be dominated by philosophy, and become a matter for scientific investigation. Just what will end up constituting "knowledge of language" and "innateness" will be determined by those empirical results. From the evidence so far in hand, it is clear that our mastery of language is to some extent the result of our innate abilities. The jury is still out on which particular theory is the more likely. I don't think that there is enough empirical evidence available to suggest that any one theory is better than the others.
Grice outlines a hierarchy in the meaning of what speakers convey by their utterances. The first level is one of conventional meanings associated with the words employed. If I say "the ball is red", then the conventional meanings associated with the words I utter constitutes the semantic contribution to what I wish to convey by that utterance. The second level is one of conventional implicature -- the implications conventionally associated with the words employed. If I say "the ball is round but soft", there is a convention associated with the word "but" implying a notable contrast. (There are, naturally, a lot of words that carry with them some conventionally associated implication in addition to the dictionary/conventional meaning of the word.) This level of conventional implicature is also, according to Grice, part of the semantic contribution of what I convey by my utterance. I choose words with which to convey what I wish to convey based on my own knowledge of the conventional meanings and conventional implicature of those words. Where my own knowledge of the conventions involved is faulty, I will tend to choose the wrong (or at least not the best) words.
The third level of Grice's hierarchy is one of first-degree conversational implicature -- the meaning of my words within the context of utterance. If I reply to your question of whether I am going to the party tonight with "I have to work", then by using the Principle of Cooperation, you can infer that I am conveying to you that I am not going to the party (among other things). The fourth level is one of second-degree conversational implicature -- the meaning of my words within context tout court. If my response to your question about the party tonight was "I like parties!", then you will have to use information about my past attendance at parties (am I a party animal or not), and my tone of voice or other attitudinal cues (sardonic or sincere) to determine whether I am conveying to you that I will or will not attend the party.
To Grice, the two levels not requiring context to establish the meaning of an utterance constitute the realm of semantics. Grice relegates the two levels requiring context to establish meaning to the realm of pragmatics. Semantics is traditionally understood as the study of meaning by focusing on the relation between signifiers, like words, etc. and what they stand for. Pragmatics, on the other hand, is traditionally understood as the study of the ways in which context contributes to meaning. So by separating out the role of context as distinguishing between semantics and pragmatics, Grice is adhering to the traditional split between the two.
Although, as Kent Bach has pointed out, there is some debate about whether Grice is merely adhering to an established tradition, or was actually instrumental in establishing that tradition. Grice's context-dependence distinction between semantic and pragmatic follows the widely used distinction between conventional meaning versus meaning in use. The only other kind of semantic-pragmatic distinction evident in the literature is that between truth-conditional versus non truth-conditional meaning. Since Grice's program is not about truth-conditional meaning, that alternative is irrelevant.
The only other threat to the plausibility of the distinction Grice draws between the semantic and pragmatic contributions to what speakers convey by their utterances, is the question of whether there really is a conventional meaning to the words being employed.
Now although I have argued elsewhere that, following Davidson, there is no such thing as "a language" in any philosophically interesting sense (only in a sociologically delineated sense), for any occasion of communication, there must necessarily be an utterer and an intended audience. If communication between the two is to be even marginally successful, there must exist a reasonably commonality in the meanings associated with the words (symbols, signs) being employed. This is the essence of the sense of "convention" analyzed by Lewis. Even if the only population involved in the convention is the two people of the speaker and his intended audience. So if a speaker is going to be even approximately successful in conveying anything to the intended audience, there will have to some conventional meaning associated with the words being employed. And the same argument can be applied, mutatis mutandis, to the real existence of conventional implicature.
So Grice distinguishes between the semantic and the pragmatic contributions to what speakers convey by their utterance, on the basis of the necessity of employing context to determine the meaning intended. Utterances examined without context constitute the semantic contribution. And semantics, as intended, naturally precedes pragmatics. Utterances examined within context constitute the pragmatic contribution. This distinction is in line with the traditional distinctions drawn by numerous philosophers of various schools, between semantic and pragmatic, and so is eminently plausible.
In his 1962 book How to Do Things with Words, J.L.Austin defined three new terms that he proposed would assist in his adopted task of analyzing what it is we do with words. These terms have caught on, and are used extensively in philosophy today. The three terms Austin introduced are:
Locution - refers to the literal or conventional meaning of the words uttered. In Austin's summary, it is the act of saying something. If you ask me whether I am going to the party tonight, and I respond "I have to work tonight", then my locutionary act was to utter the words that I have to work tonight. Austin further analyzes a locutionary act as comprising a phonetic, a phatic and a rhetic act corresponding to the verbal/auditory, syntactic, and semantic aspects of the speech act.
Illocution - refers to the effect the speaker intends to achieve in making the utterance. In Austin's summary, it is the act done in saying something. By uttering the words (performing the locutionary act) "I have to work tonight" I am telling you (performing the illocutionary act) that I am not going to the party. Austin claims that to perform a locutionary act is eo ipso to perform an illocutionary act. In Gricean terms, a locutionary act focuses on the conventional meaning of the uttered words, while the illocutionary act focuses on the intended conversational implicature of the utterance.
The illocutionary force of an utterance is the speaker's intention in performing the locutionary act. An illocutionary act is an instance of a culturally-defined speech act type, characterised by a particular illocutionary force. For example, promising, advising, warning, etc. In 1975, John Searle used Austin's concept of "illocutionary force" to set up a classification of illocutionary speech acts that has also become widely accepted in the literature:
"assertives" are acts that commit a speaker to the truth of the expressed proposition;
"directives" are acts that are intended to cause the audience to take a particular action -- illocutionary acts intended to be perlocutionary acts by causing the audience to take some action;
"commissives" are acts that commit the speaker to some future action;
"expressives" are acts that express the speaker's attitudes and emotions towards the proposition;
"declarations" are acts that change the reality in accordance with the proposition of the declaration.
In 1985, Searle and Vanderveken provided a detailed analysis of illocutionary force, and documented seven features they claimed defined all possible variations of illocutionary force. For this essay, I need not go into that detail.
Perlocution - refers to the actual effect of the utterance upon the audience -- how it is received by the listener. In Austin's summary, it is the act done by saying something. In Austin's words, a perlocutionary act occurs when a speaker says something that produces "an effect on the feelings, thoughts, or actions of his audience". The "perlocution" of a speech act (my telling you that "I have to work tonight") is the way it is received by the audience -- the way in which you understand the meaning of what it is I utter. In Grice's terminology, the "perlocution" of my utterance is how the audience understands the utterance, including all conventional and conversational implicature of my utterance. My illocutionary act is to tell you that I am not going to the party tonight by performing the locutionary act of uttering "I have to work tonight". But my speech act has a perlocutionary component only if you understand my intent, and accept that I will not attend the party tonight.
As a further example, consider the actions involved when I yell "Low Bridge!". My locutionary act is to utter the words "low bridge". My illocutionary act is to warn you to duck. My perlocutionary act (if successful), is to cause you to duck. Austin makes it clear, however, that it is not necessary for the consequences produced by an utterance to be intentional for there to be a perlocutionary act involved in the utterance. If my utterance was instead "Gee! Look at that beautiful bridge!", and you still ducked, then I still performed the perlocutionary act of getting you to duck -- even though that was not my intention.
The distinction between the illocutionary act and the perlocutionary act is the distinction between my (the speaker's) intention, and your (the audience's) response to an understanding of my intention. A perlocutionary act is one that is the cause of an effect on the audience. But notice that in defining perlocutionary acts as involving an effect of a cause, Austin is here using "cause" in a sense different from the normal physical sense. Austin claims that there are two senses of "cause", a physical and a non-physical sense. It is the latter which is relevant to his distinction between illocutions and perlocutions. He is thus using the notion of "cause" in its customary or sociological (arguably, the original) sense. In this sense of the word, that which is caused is the free and deliberate act of a conscious and responsible agent. "Causing" you to do something means affording you a motive for doing it. The argument I voice to support my position is an illocutionary act. My convincing you (if I do) is my perlocutionary act.