I'm replacing symbol with sign for this topic. A symbol has to be learned (http://anthropology.berkeley.edu/sites/default/files/43.pdf (uncorrected proof but URL suggests no error in view of poster/author, prof., biological anthropology)), but it's just one kind of a sign and a sign does not necessarily have to be learned. A fetus may not be learning symbols for all of what it needs to signify but nonetheless could have signs for its signifieds. It might invent signs or they might be biologically (viz., genetically) formed.
What the fetus would need to signify is a referent or object (http://open.lib.umn.edu/communication/chapter/3-1-language-and-meaning/). What the fetus could choose (if and when fetal choice is possible) is thus a referent or object.
Once a referent is known, in support of immediacy of a sign, this applies: "An object, in and of itself, has no meaning without a sign being interpreted into a pragmatic context in the mind. . . . '[S]igns enable us to transform objects . . . into meanings[.]'" (.) It is also said that the sign refers to a conceptual referent which in turn refers to a possibly physical thing, the referent in that sense being a thought. (, by assoc. prof. George Fowler.) If that's how we should understand the issue, then a fetus that can make a choice necessarily can have a thought. That may raise a question of what a thought is, at the most elemental level. It is relevant that animals think (as mentioned above); that's known simply from the basis that their survival would be impossible without it, although the constituents of thought in animals may be less than in humans (https://news.harvard.edu/gazette/story/2002/03/scientists-think-that-animals-think/), at least for adult humans.
A thought or meaning being nonexistent is equivalent to the object itself being nonexistent as far as any perceiver would be concerned. However, at least one object exists as far as the the perceiver would be concerned and that means that the fetus (as perceiver) would have information, at a fundamentally simple level, about the object and therefore would have a basis for having a meaning or thought about the object. More from Fowler, id. (italicization omitted): "Commonsense tends to insist that the signified takes precedence over, and pre-exists, the signifier: 'look after the sense', quipped Lewis Carroll, 'and the sounds will take care of themselves' . . . . However, in dramatic contrast, post-Saussurean theorists have seen the model as implicitly granting primacy to the signifier, thus reversing the commonsensical position. . . . Commonsense suggests that the existence of things in the world preceded our apparently simple application of 'labels' to them (a 'nomenclaturist' notion which Saussure rejected . . .). Saussure noted that 'if words had the job of representing concepts fixed in advance, one would be able to find exact equivalents for them as between one language and another. But this is not the case' . . . ."
An object might be known or not by an organism. For example, an oyster might or might not know about a certain grain of sand nearby. If it is known (e.g., the grain of sand irritates the oyster), the existence inside the oyster of that knowledge means the knowledge is the sign in the oyster. The oyster thereby has a sign to represent the irritating grain of sand. The oyster might make a secondary sign that may be more efficient and possibly more abstract than the knowledge itself, but it had a sign commencing when it acquired the knowledge and is never without a sign as long as it knows of the grain. It evidently knows of the irritating grain, since it responds to the grain by making it into a pearl. This may be the key that I had missed in earlier posts: because a choice known to a fetus is a subject of knowledge within the fetus, the knowledge itself is a sign. The fetus may develop a better sign, but, as long as it has knowledge of that choice, it is never without a sign for it. One writer considered a signified as part of a sign, not simply something the sign referred to; perhaps this is the same idea. The knowledge itself being a sign means the sign is known to the organism as soon as the knowledge is within the organism. Thus, there's no delay from possession of the knowledge until possession of the sign.
Plants, which were discussed above, communicate with each other and have memory with which they can encode, retain, and recall, although they probably don't think. (.) The issue of whether thinking requires a brain was, I think, discussed above. "But [per ibid.] you don't need neurons in order to have cell to cell communication and information storage and processing . . . . [and, e]ven in animals, not all information is processed or stored only in the brain." Plant communication can be internal to one plant; "[d]ifferent parts of the plant communicate with each other". Ibid. I do not think that a fetus's thinking requires that it have a brain, partly because fetal thinking may be less complex than that in adults and other fetal structures may suffice to process its information. (Information theory allows for information to be very simple; a single bit, representable by a base-2 number as only 0 or 1 (thus true or false, on or off, etc.) is information.)
Two confusions should be addressed. One line of discussion is that a sign includes both a signifier and a signified, whereas it would be clearer to treat a sign as a signifier and to have it refer not to a signified but to a referent. And "object" and "referent" are often treated interchangeably, but among many people "object" conjures physicality or at least reality (e.g., whether a Martian is an object when Martians don't exist), so "referent" is likely a better word.
A couple of arguments can essentially be dismissed. An argument (stated by Fowler, id.) that a sign may require a collection of signs in order for the one to be understood is trivial for this topic, since the first set of signs possessed by a fetus would consist of at least two (saying there's only one choice is metaphorical, there having to be at least two choices for choice to exist, each needing a sign). Also, an argument (also stated in ibid.) that one referent can have more than one sign (e.g., synonyms) and one sign can be for more than one referent (e.g., with a pun) would not apply to this topic, since a choice between two choices, leading to a choice between two signs, requires that each immediate choice have an immediately unique sign. You might have two warm coats and a thin sweater, but if you're about to freeze and need to choose a coat, you'll distinguish between the coats.
Hypothetical examples of cases in which referents are known but no signs exist for them are welcome. I haven't thought of any.
(All URLs were as accessed Nov. 12, 2017.)