Indeed, Wikipedia repeats this assertion:
"Sometimes onomatopoeic words can seem to have a tenuous relationship with the object they describe. Native speakers of a given language might never question the relationship; however, because words for the same basic sound can differ considerably between languages, non-native speakers might be confused by the idiomatic words of another language. For example, the sound a dog makes is bow-wow (or woof-woof) in English, wau-wau in German, ouah-ouah in French, gaf-gaf in Russian, hav-hav in Hebrew, wan-wan in Japanese and hau-hau in Finnish."I would argue that in fact that is not evidence of a tenuous relationship. Rather, the cross-language similarity of these onomatopoeias is evidence that they do in fact represent the underlying sounds.
What, then, to make of the differences?
We need to realize that animals, rivers, and hammers do not "speak" the same way that humans in their proximity speak. They make sounds, which humans parse. However, while human speech can be divided into phonemes, animal sounds cannot generally be divided into those same phonemes.
A parallel could be drawn to the way Hebrew and Aramaic developed daled and zayin in opposition. That is, in some words, both Hebrew and Aramaic have zayin. In some words, both Hebrew and Aramaic have daled. But in some words, Hebrew has zayin while Aramaic has daled. For example, Hebrew zu vs. Aramaic di, Hebrew zahav vs. Aramaic dehav. The cause for this, one prominent suggestion goes, is that Proto-Canaanite had three letters, zayin, daled, and some in-between letter that sounded like the /th/ in "either." When they adopted the Assyrian alphabet, there were only the symbols for daled and zayin. What, then, to do with this additional letter? Hebrew mapped it to the zayin while Aramaic chose to map it to the letter daled. (The same is true for other distinctions between Aramaic and Hebrew, such as the tav/shin switch-off.)
Does this mean that the Hebrew zahav or the Aramaic dehav have only a tenuous connection to the original word as pronounced? No. Rather, given a limited orthographic inventory, each language chose to map it to a letter that approximated the sound.
Similarly, language speakers have a limited phonemic inventory, a set of sounds they make when pronouncing the language. When one maps a sound outside of that phonemic inventory to something within the phonemic inventory, of course he makes an approximation. No one says that a dog says "bow-wow" in exactly the same way that a human does. It is an approximation of the sound.
And, just as different languages approximated the /th/ sound to different letters, on occasion different languages approximated the sounds in a dog's bark using different phonemes.
Furthermore, speakers of different languages have different phonemic inventories. Speakers of certain Asian languages do not distinguish between the /l/ and the /r/. The Hebrew resh rolls, whereas the English "r" is a hard r. Modern Hebrew does not have a "w" as a consonant.
Can we then complain that Hebrew does not have the dogs say "woof-woof" if there is no "w" in Hebrew? Instead, it maps that sound to a heh.
Alright, hav-hav is actually a fairly old onomatopoeia, and perhaps back then they actually did pronounce a vav as a "w." Still, who says that the first sound is strictly either an /h/ or a /w/? Perhaps it is something in between, just as some people pronounce both the /w/ and the /h/ in words such as "when" (and which in Old English was hwenne, with the h first). Perhaps different languages picked up different aspects of the sounds being produced.
Different features might influence the particular choice. For example, /f/ and /v/ are related, the former being unvoiced and the latter voiced. /v/ is related to /b/, one being fricative and the latter being plosive. Perhaps at the beginning of a sound, to try to approximate the sudden explosion, some languages might lean towards the plosive over the fricative. And the examples given above showing these distinctions between languages are misleading, a) because the differences are either due to different spelling rules of sounds, or b) because the phoneme corresponding to that letter (grapheme) in that language is not the same as it would be in English, and thus the phonemic difference is much less. For example, German "w" sounds like English "v."
Particular meanings within these languages might slightly shift the pronunciation within the language. Thus, hav-hav means "give, give" in Aramaic (from יהב) and a cock, or rooster, says "cock-a-doodle-doo."