This page is still a big work-in-progress. The code, in particular, is a mess right now because I exported all this from an .odt file into an .html. Not the best shortcut to attempt, it turns out.

Introduction

This webpage serves as an attempt to uncover the procedure used to create all the lines in ICO (both used and cut) that utilize the hieroglyphic script seen above.

The fandom community surrounding Fumito Ueda’s work commonly refers to this language as ‘The Runic Language’ (or ‘Runic’ to be concise). However, this is not an official term as far as I can tell, and the developers have usually referred to it as “Yorda’s Language”.

At present (1/21/26), this is what is listed on the Team ICO Wiki (which denotes our community’s collective understanding) regarding the process of decoding this language:

Translating the language manually

1. Translate the runes into their corresponding letters to get a word in backwards Romaji.

2. Flip the order of the letters to form the Romaji word. The tricky part here is that a few vowels will often be missing; one will need to figure out what they are and where they go in order to get a real word. (This was an intentional linguistic move by Kuwabara to make Yorda's speech seem more delicate and ethereal.) Fortunately, Yorda's speech is very basic, and with a bit of linguistic common sense, it's relatively easy to find where the vowels were taken out.

3. Once the final Romaji word is discovered, translate it using whatever means desired to get the word in English.

This portion’s words have been almost entirely untouched since the Runic Language page’s creation in 2009. It is incredibly non-specific. But the community has been satisfied by it for all this time, nonetheless. Through my findings, I hope to change that.

You will see my observations and notes first, before you see the script lines I picked apart. I’ve organized things this way to help orient readers, and show the patterns you can latch onto as you read.

Number references (e.g. See: 00) apply to the script section’s File Numbers, which categorize each line.

Observations

Repeated Terms (Exception of Yorda’s name)

The following are terms with fairly consistent transformation, and spelling, within the language.

naze = ezn (See: 32, 48, 59) [100% consistent]

saa = ahs (See: 31, 45) [100% consistent] (middle ‘a’ consistently transforms to ‘h’ via a → h substitution)

oide = dio (See: 31, 45) [100% consistent]

daijoubu = ad boju (See: 77, 81) [100% consistent]

kaasan = nsk (See: 46, 61) [100% consistent]

waka(-ranu or -ranai) / waka(-tteru wa) / waka(-tte) = arkw / wur tkw / e akw (See: 48, 53, 58, 67) [100% consistent]

The ‘kw’ of ‘waka’ is consistently left alone. While different forms of the verb change the rest of the term.

hontouni moshiwakenai = ynst (See: 50, 79) [100% consistent]

-Pronouns-

omae (wa / wo / ga) = awm / owm / agm (See: 54, 55, 57, 66, 69) [83% consistent] (See Outlier: 49 (This is a unique instance of w → h substitution for this term)).

The ‘m’ of ‘omae’ is consistently left alone and then paired with its follow up particle (be it wa, wo, or ga).
We see a similar case with ‘anata’ and ‘watashi’. Making it a potential rule of all pronouns.

anata wa = aht n (See: 9, 46, 61) [100% consistent] (‘wa’ consistently transforms to ‘ha’ via w → h substitution)

One instance of ‘anata’ was spelled ‘htna’(See: 21), potentially making it less consistent. However, it’s unclear if the original draft really included the ‘wa (は)’ particle or not. The term could be paired with the start of ‘hitori’ to explain the ‘h’.

watashi (wa / wo / no) = aws atw / ows atw / onhs atw or just ‘atw’ when without particle (See: 53, 56, 59, 60, 62, 63, 68, 79) [80% consistent] (See Outliers: 20, 50)

-The English “Yes”-

Terms in Japanese that mean “yes”, tend to become the actual English word, reversed in Yorda’s Language: sey. These terms include ‘ee’ (ええ) (See: 68) and ‘un’ (うん) (See: 83).

Substitutions

When writing in Yorda’s Language, one can substitute certain letters for others. This allows the limited Romanization of Japanese to include all 26 English letters, while also diluting the meaning of words.

A – H (See: 31, 45)

CH – T (See: 52, 53, 61)

E – A (See: 72, 76)

H – N (See: 56) or W (See: 49)

I – Y (See: 49, 50-50, 52, 56, 59, 60, 64, 65, 67, 70, 74, 79) U? (See: 62, 77, 81)

K – Q (See: 29, 47, 49, 52, 53, 57-57, 62, 67, 68, 74) or C (See: 20, 50, 56, 57, 62, 69) or X (See: 54, 86)

N – M (See: 46) or U (See: 60, 68)

O – U (See: 69, 72)

R – L (See: 57, 64, 68 [Excluding instances of Yorda’s name]) or S (See: 9)

SH – X (See: 50)

W – H (See: 9, 46, 49, 61)

? – P (See: 47)

? – V (See: 47)

Potential Reasonings:

-Japanese Pronunciation/Association-

CH → T

‘CH’ is the softer alternative to the dental ‘T’ sound. And Japanese speakers, over the course of centuries, dropped the more precise ‘ti’ in favor of the softer ‘chi’. If you look at any syllable chart for the language, you will notice other consonants are often paired cleanly with their vowels (na, ni, nu, ne, no) but T is not one of those cases (ta, chi, tsu, te, to).
‘Team’ (チーム) is a good example of a loan word’s ‘ti’ (or ‘tee’) sound becoming ‘chi’.

H ↔ W

The particle は (ha) has a special use case in Japanese. When used to indicate the topic of a sentence, it is instead pronounced ‘wa’ without changing the character used (typically, ‘wa’ is associated with this character: わ). This may be why the letters are allowed to swap places in File 49.

I → Y

In English loan words that end with ‘Y’, often creating the ‘ee’ sound, Japanese speakers use syllables that end in ‘I’, which for them, is pronounced just the same (‘ee’). ‘Lucky’ (ラッキー) and ‘Canopy’ (キャノピー) are exmaples of this.

K → X
To pronounce English loan words such as ‘Expo’ (エキスポ) or ‘Box’ (ボックス), Japanese speakers will use the ‘ki’ or ‘ku’ syllable (with the i and u’s pronunciation devoiced), in tandem with ‘su’, to sound out the ‘X’ (ksu). Thus creating an association with the letter K.

K → Q

‘K’ syllables are used for English loan words featuring Q, such as ‘Quest’ (クエスト) or ‘Quark’ (クォーク).

K → C

As with the entry above, ‘K’ syllables are used for many English loan words featuring C, such as ‘Capsule’ (カプセル) and ‘Cake’ (ケーキ).

R → L

“Japanese only has a singular liquid phoneme /r/ which is usually pronounced as an alveolar tap [ɾ], but can also be pronounced as an alveolar lateral approximant [l]. This is in contrast to English, which has two liquid phonemes, /r/ and /l/, usually pronounced as the postalveolar approximant [ɹ̠] and the alveolar lateral approximant [l] respectively. As a result, when translating names into English, especially fictional names that are intended to sound foreign, either R or L can be used.”
– Team ICO Wiki (Yorda article)

SH → X

This is more of a Chinese association, but due to the undeniable influence of the Chinese language on the Japanese language, it should count among the Japanese associations, I think. X, as written in Chinese, is pronounced with a sound comparable to the ‘SH’ sound in English.

-Similarity of Form in English-

e → a / n → u
These letters are diagonal reflections of each other in their lowercase forms.

H → N (h → n) / N → M (n → m)
These letters bear similarities in their structure, for both upper and lowercase.

A → H
These letters bear similarities in their structure, in uppercase form.

O → U

An O could be perceived as a ‘closed’ U? Uncertain.

r → s

Due to their hook-like quality, perhaps these letters were considered structurally similar in their lowercase form? Uncertain.

Yet Unclear Reasonings: I → U

Revised Lines

This is not a rule of the language, so much as it is a consequence of development that becomes an obstacle to us, as decoders.

A few of The Queen and Yorda’s lines, constructed at an unknown point in time, were revised in structure and vocabulary for the game’s NG+ captions (first included in the Japanese release, on December 6th, 2001).
Because we rely so heavily on the NG+ translation feature to guide our understanding of Yorda’s Language, these revisions inadvertently muddle the original meaning of the hieroglyphic captions.

Luckily, the revisions are not often drastic. Most involve a few words being changed into synonymous terms that either carried slightly different implications or sounded better to the developers. But overall, the original spirit of these lines remained intact.

Even so, it puts us in the difficult position of having to guess whether a line was encoded by an unknown rule, or if what we see are actually the remnants of an earlier draft. Some remain unclear even now.

Transcript and Supposed Process

(Note: I am a novice when it comes to Japanese, so identifying where the spaces should be placed in the Romaji transcription was difficult at times. I’m sure I’ve made mistakes, so please feel free to double check my work.)

File 9:

anata dare?

doko kara haitte kita no?

[anata da{r}e]

[doko] [kara] [haitte] [kita no]

Glitterberri suggests earlier drafts used:
あなたは

[anata {w}a]

esad aht n

okd ar eti on

File 16:

yoruda...

yo{r}uda

yld

File 20:

watashi wa koko wo

hanarerarenai no...

[watashi wa] [ko{k}o wo]
[hanare][rarenai no]

aw stw ock

erh nia era

File 21:

anata hitori de itte

[anata hito][ri de] [itte]

htna dir eti

File 29:

ano hito wo
okorasete shimatta wa...

[ano] [hito wo] [o{k}orase] [shimatta wa]

on wti e

srq atms

File 31:

saa kaette oide yoruda

[s{a}a] [kaette] [oide] yo{r}uda

ahs etk dio

yld

File 32:

yoruda

naze damatteiru no da?

yo{r}uda

[naze] [damatte][iru no da]

yld

ezn e amd nur

File 45:

saa oide yoruda

[s{a}a] [oide] yo{r}uda

ahs dio yld

File 46:

anata nanka

watashi no kaasan janai wa

[anata] nanka

watashi no [kaasan] [? ja][{n}ai wa]

Glitterberri suggests earlier drafts used:
あなたは

[anata {w}a]

Mystic suggests an earlier draft used:

ほんとうに in place of ‘nanka watashi no’ and had こと placed where the ? is
[hontou ni] [koto]

aht n nu otn

nsk ajk wiam

File 47:

kikiwakenonai ko da...

[{k}ikiwa keno][{n?}ai ko da]

onk wkq adpv

File 48:

naze wakaranu

[naze] [wakaranu]

ezn arkw

File 49:

omae wa soto no sekai deha ikite wa ikenai no da yo

[omae {w}a] [soto no sekai] [de{h}a] [i{k}ite wa {i}kenai no da yo]

ahm iks ont we

ydn ank ytq

File 50:

kitto hidoi koto wo iwareta no ne
watashi no sei de gomennasai...

[{k}itto] [hidoi koto wo] [{i}wareta no ne]
[wata{sh}i no] [sei de] gomennasai

Mystic suggests earlier drafts used:

所為で 私 本当に申し訳ない in place of ‘watashi no sei de gomennasai’

[seide wata{sh}i hontouni moshiwakena{i}]

tic ow dh rwy

ynst xtw dies

File 52:

zuibun to machikutabireta yo

[zu{i}bun to]

[ma{ch}i{k}utabireta yo]

onu byz ytr

batuqitam

File 53:

yoruda watashi no kimochi mo
wakatte okure

yo{r}uda [watashi no] [{k}imo{ch}i mo]
[wakatte okure]

yld onhs atw

mt mq e akw

File 54:

omae wo kurushimeta kunain dayo

[omae wo] [{k}urushime taku][nain dayo]

owm kt msx yda

File 55:

konnani omae no koto wo
omotte irunoni

[konnani] [omae no koto wo]

[omotte irunoni]

ian nk owm

nnur etm

File 56:

sonna watashi wo oite

doko he ikou toiu no dai?

sonna [watashi wo] [oite]

[do{k}o {h}e i kou toiu no da{i}]

ows atw eti

yduk incd

File 57:

omae wa koko deshika

ikirarenai no dayo

[omae wa] [{k}o{k}o de shi{k}a]
[ikira{r}enai no dayo]

awm aqhs

edcq nlrk

File 58:

wakatteru wa...

[wakatteru wa]

wur tkw

File 59:

deha naze watashi no soba ni inai

[deha] [naze]

[watashi no] [soba ni ina{i}]

ahd ezn onhs

atw yni abs

File 60:

watashi wa modoranai

[watashi wa] [modora{n}a{i}]

aws atw yurom

File 61:

kaasan

anata wa machigatteru wa

[kaasan]

[anata {w}a] [ma{ch}igatteru wa]

nsk

aht n rt agtm

File 62:

watashi wa jibun no

ikitai youni ikiru

[watashi wa]
[j{i?}bun no] [i{k}itai youni] [i{k}iru]

aws atw

obuj nytq rc

File 63:

sono daishou toshite watashi no

inochi ga ushinaware you tomo

sono daishou toshite [watashi no]

inochi [ga] ushinaware you tomo

utt onhs atw

ag ar omdnh

File 64:

tsuminonai shuzoku nogisei noueni

ikinagaraeru yori zutto mashi dawa

[tsumino]?[nai shu?zoku] [nogisei noueni]

[ikinaga{r}aeru] [yori] [zutto] [mash{i}] dawa

omuwkzyus eisg

rl agn ry otzym

File 65:

nani wo iidasu no da? yoruda

[nani wo] [i{i}dasu no da] yo{r}uda

owi anuy yld

File 66:

arehodo sunao datta omae ga...

arehodo [sunao datta] [omae ga]

Glitterberri suggests an earlier draft began with:

そんなに

[sonna ni]

i an td nus agm

File 67:

koreda keitte mo

wakaranai no kai?

[koreda kei][tte mo]

[wakaranai no {k}a{i}]?

ekd er omt

yq nia arkw

File 68:

ee
watashi no kimochi wa kawaranai...

[yes] [wata shino]
[kimochi wa] [{k}awa{r}a{n}ai]

sey onhs

atw awmi ulwq

File 69:

omae wo sokomade kaetano wa

sono tsuno no haeta kodomo kai?

[omae wo] [soko made] [{k}aetano wa]

[sono] tsuno no haeta [k{o}domo kai]

Mystic suggests earlier drafts used:
いけにえ rather than ‘tsuno no haeta’

[ikenie]

owm edm ok wntc

nos enk omduk

File 70:

kono ko wa kankeinai wa!

[kono ko wa] [kankeina{i} wa]

aw knk aw ynknk

File 71:

izure niseyo sono ko niwa

sukoshi oshioki wo shinai to ne

(supposedly laughter. ‘fu fu fu’, unsubtitled in the Japanese.)

[izure ni seyo]

fff
oye neruz

File 72:

[continuing from above]*

[s{o}no ko niwa]

sukoshi [oshioki wo shinai to n{e}]

awnk onus

ant ani ikis

*File 72’s first play-through captions are a continuation of the sentence that began in File 71.
However, the ‘File 72’ found amid the New Game Plus captions provides a new sentence that wasn’t actually processed into Yorda’s Language.

File 74:

saa jamamono wa inaku natta

issho ni kaeruu yoruda

saa →yo{r}uda
[jamamono] [wa inaku natta] [issho n{i}] [{k}aerou]

Mystic suggests an earlier draft began with:

去れ

[sa{r}e]

las yld

n amj tuk ys orq

File 76:

yamete!

[yam{e}te]

etam

File 77:

daijoubu?

[da][{i?}joubu]

ad boju

File 79:

gomennasai

watashi no sei de konna kotoni

[gomennasai

watashi no] [sei de] [konna kotoni]

Mystic suggests earlier drafts used:

本当に 申し訳ない 私 本当に 申し訳ない

[hontouni moshiwakena{i}] [watashi hontouni moshiwakena{i}] [de] [ru kotoni]?

ynst ynst

atw ed intk ur

File 81:

inaiyo kiechatta…

mou daijoubu yo

[inaiyo] [kiechatta]

[mou] [da][{i?}joubu yo]

yni at ek

um ad boju

File 83:

un...

[yes]

sey

File 86:

Ato mou sukoshi yo

[Ato mou] [su{k}oshi yo]

um oty sx

File 91:

arigatou…

arigatou

nonomori

nnmr

File 112:

sayonara

[sayonara]

arn oys

Final Notes

The means of both constructing and reverse engineering Yorda’s language are far more complex than many have been led to believe.

It was said in the ‘Walking with Giants’ interview (found among the bonus features of the ICO & Shadow of the Colossus Collection), that those on ICO’s planning team used to call the process of converting the script to Yorda’s language: Self Conversion (自分変換). This moniker is faintly elaborated on by Kenji Kaido with the following translated quote:

“Each person has developed his own individual conversion rule based on their feel and experience.”

The planners included Junichi Hosono, Kei Kuwabara (responsible for the majority of the language’s hieroglyphic visuals), Tsutomu Kouno, and, of course, the duo of Kenji Kaido and Fumito Ueda. So this quote implies a minimum of five overlapping rules, one created by each individual.

However, it is difficult to accredit individual rules, and whether these rules were applied evenly across all the dialogue.
For instance, would the many details of the letter substitution system count as one rule, or several? And why are substitutions absent for some lines, but not others? Was the absence a part of an existing rule/pattern, or was it simply because a planner was not present to use their conversion method at the time of the script addition/adjustment?

In truth, there may be no rhyme or reason to some of these decisions, beyond what appealed to the developers at the time.

But even so, a better understanding of these conversions is possible. At time of writing, I have marked 33 lines as completely understandable transformations. This means we can provide a 100% plausible (to our limited knowledge) means of transformation to approximately 76% of all the lines that use Yorda’s Language. And many other lines are very close to meeting that standard of completeness.