Old Japanese is commonly reconstructed as having the following thirteen consonants: /p b t d k g s z h m n r w j/. The voicing exhibited in consonants is typically analyzed as having historically stemmed from intervocalic voicing, meaning Proto-Japanese (also called Proto-Japonic and Proto-Japanese-Ryukyuan) would likely have lacked /b d g z/ as distinctive phonemes. But a few problems arise when establishing a proper Proto-Japanese reconstruction due to bad comparisons with Ryukyuan languages. Namely, /w/ and /j/ are sometimes regarded as having stemmed from */b/ and */d/ respectively, and /h/, which should really be /ɸ/, is almost unilaterally considered to have originated from */p/.

Origin of *w

The first one originates from comparisons with the Southern Ryukyuan languages, which include Miyako, Yaeyama and Yonaguni. In these languages, words that are normally pronounced with a /w/ in Japanese are pronounced with a noticeable /b/. Because Ryukyuan languages tend to be much less influenced by Chinese, and these languages are spoken on the islands furthest from central Japan, this is regarded as evidence that */b/ was the likely origin. Unfortunately, this very idea is quite false and fails to consider two things. First, while it is true that the core vocabulary of the Ryukyuan languages tends to feature less words of Sinitic origin, this bears no relation to phonology, and the actual phonology of the Ryukyuan languages is in fact quite heavily warped. Second, evidence from the dialects themselves suggest a number of discrepancies:

ban 'I, myself' (:: NR wan, J wa)
bata 'stomach' (:: NR/J wata)
barazan 'knotted rope system' (:: NR/J warazan)
baarun 'to laugh' (:: NR waarun/warain/wareen)
binga 'man' (:: NR wikiga/yikiga/yinga)

yarabiiyarabi 'childish' (:: NR warabi, J warabe 'child')
kaa 'water well' (:: J kawa 'river', Takara & Kuroshima Islands kaa 'well')
ya 'topic particle[/i] (:: J wa, Kagoshima a~ya~wa)

In the vocabulary list above, it is clear that a correspondence exists between Southern Ryukyuan /b/ and Northern Ryukyuan /w/ as well as Japanese /w/. For example, the root *ba- 'I, myself' in the south is *wa- in the north. However, what needs to be pointed out is the very limited position of this correspondence: /b/ only surfaces for /w/ in word-initial position. So this begs the question: why? If */b/ were truly the source, why does it not appear elsewhere, and why does it actually appear to elide and assimilate? It seems a little far-fetched to posit *kaba > kaa 'well', *barabi(i) > yarabi(i) 'child', *ba > ya 'topic particle', etc. In fact, if we posit that /b/ became /w/ which then elided in medial position, why does the second /b/ still remain in warabe? Another question to wonder is why would */b/ spontaneously shift to /j/ in some words and the topic particle itself? Why is it that we observe /ba/ for the object marker instead?

There are simply too many discrepancies to legitimately posit that /w/ was originally /b/. Instead, by considering */w/ as the proper origin, the processes of intervocalic elision and gliding become clear, and parallels to these phenomenons can be found in Kagoshima. To explain the gliding of /w/ from /j/, we merely have to consider the fact that this change appears most prominently around front vowels. The initial /j/ in the Miyako word yarabiiyarabi 'childish' can simply be attributed to phonological parallelism due to reduplication leading the middle /w/ to be placed beside the front vowel /i/. The initialism of */b/ can also be attributed to the fact that these dialects exhibit fortition of initial plosive consonants. Yonaguni, for instance, observes the following changes as well: /j/ > d, /tɕ/ > /t/, /ɕ/ > /tɕ/, etc.

However, if this change is secondary, then we also have to explain why words like washiki 'weather' in Yonaguni are permitted. In this case, washiki luckily has a parallel form in Okinawan: qwaachichi. The Okinawan form provides evidence that the historical word started with *owa-. So, if we consider that fortition does not affect sounds in word-medial position, then washiki is perfectly reasonable assuming the initial vowel */o/ was raised and later clipped.

Origin of *j (*y)

In the same way that /w/ was thought to originally stem from */b/, /j/ was proposed to have originally been pronounced */d/. This idea stems not from Southern Ryukyuan as a whole, but instead, it solely rests on one variant: Yonaguni. In Yonaguni, words like standard Japanese /jama/ 'mountain' become /dama/. The name Yonaguni itself is pronounced /dunaŋ/ in the regional language.

duru 'night' (:: J yoru)
dumi 'wife' (:: J yome)
da(a) 'house' (:: J ya, ie)

It seems a tempting idea to believe that because Yonaguni is the furthest of the entire Ryukyu Islands, its language is therefore the closest to the original language. But consider the following examples as well:

aidi 'sign' (:: J aizu, OKN aiji)
kadi 'wind' (:: J kaze, OKN kaji)
daidubu 'alright' (:: J daijoubu)
barazan 'knotted rope system' (:: NR/J warazan)

The data set above suggests another relation between Yonaguni /d/ and Japanese /z/ and /dj/. Here, it seems that, like all other Ryukyuan languages, Yonaguni underwent vowel raising of /e/ to /i/, causing kaze to become kaji, and vowel fronting of /u/ before alveolars, causing aizu to become aiji. The phoneme /dʑ/ then underwent fortition, becoming /d/.

It seems difficult to try and explain how the proto language's */d/ would have variously become /d/, /j/, /z/, /dj/, /zj/ in its daughter languages. This analysis would also require us to explain how /di/ became Ryukyuan's /dʑi/ which became Old Japanese's Yotsugana syllables: [dzu], [zu], [dʑi] and [ʑi]. We therefore have to rule out this theory as being plausible.Therefore, the change /dʑ/ > /d/ is the only other possibility and was localized to Yonaguni.

Going back to /j/, another discrepancy seems to arise: morphemes which are /d/ in initial position, like da(a) 'house' (Japanese ya), are /j/ in medial position: chimunuya 'kitchen' (lit. 'house of the kitchen'). Here, the earlier parallel with /w/ and /b/ is uncanny. If */d/ as the source is not plausible, then fortition of /j/ in initial position is the only possibility. If we assume that /j/ became /dʑ/ at some point in the language's history, then the jump to its modern form /d/ is a relatively small feet.

Origin of *h

Last on the list, Japanese's modern /h/ is almost always reconstructed as having originated from */p/. The reason for this belief is similar to the other two: a few of the Ryukyuan languages both in the north and the south have /p/ where modern Japanese has /h/. A second piece of evidence reinforcing this idea is the fact that reduplication in Japanese, which is associated with voicing, causes /h/ to become /b/: hitobito 'people'. It is also worth noting that strengthened syllables that would otherwise contain /h/ gain the sound /pp/.

We know from historical transcriptions of Japanese into Portuguese and Russian that the h-series in modern Japanese was once pronounced /ɸ/. Okinawan also partly retains the pronunciation /ɸ/. So while positing */p/ to */h/ might be a stretch, positing lenition of */p/ to /ɸ/ and then /h/ does not seem all that far-fetched. However, if we consider that fortition is prominent in Ryukyuan, then the fortition of [ɸ] to [p] is also not unlikely, and this notion fits well with the initial-medial difference exhibited with the other consonants.

It is also important to note that all the Ryukyuan variants that exhibit /p/ where Japanese has /h/, also exhibit other forms of fortition. Moreover, this correlation seemingly only occurs in word-initial position, never in word-medial position, which ties in quite well with the other fortition-based sound changes in Ryukyuan. Considering this, I believe that skepticism should be observed when considering the Ryukyuan data.