what conditions truth?

A mango is neither true nor false. The taste of a mango is neither true nor false. An understanding, statement, or belief about the mango's taste CAN be true or false.

I personally believe in thought outside of language. I believe it is possible for Seb to taste a new fruit without immediately assigning it linguistic labels. But the moment he says, "A mango tastes like X," to anyone OR to himself, he has made a comparison that relies on linguistic categories. to determine whether or not his belief is true when he tastes another mango, he must communicate with himself over time. for his assessment of the taste to be judged "true" or "false," it must be expressed (in language or in signs).

if seb tastes the mango and never states his belief at all, then tastes another mango and recognizes the same taste, i would argue that he has not made a prediction, but has simply experienced the same thing twice. admittedly, these distinctions are pretty crude, but what i'm really trying to get at here is the fact that "true" and "false" are concepts built up in language.

let's return to the sky for a second. imagine trying to express "the sky is blue" without using words. suppose i point to my skirt (which is also blue), and then indicate the sky. setting aside for a moment the fact that it would be difficult to understand that my pointing indicated the colors, the problem we run into becomes even more severe. my skirt and the sky are both blue, but they're not at all the same color. we only recognize them both under the same label ("blue") because we already know that "blue" encompasses many different shades and hues of blue. without the already-defined totality that the linguistic label provides, there is no way for me to describe the color, even by comparison.

this example also demonstrates the extremely useful nature of words that define totalities like "sky" and "blue". without them, we'd be stuck trying to agree on definitions every time we needed to refer to them. in this framework, confirmation holism (which wikipedia article Seb linked to in his comment) is right on. but it's my belief that much of the framework from which we assess concepts already exists in the language we use to talk about it. saying "'the sky is blue' is true" takes for granted that 'the sky' is a well-defined noun (which doesn't include the air between my eyes and 'the sky', for example), the fact that 'blue' is an already-defined color that includes many shades, and the idea that the 'truth' of a statment is a simple property.