Na-Adjective vs. Noun in Japanese: The Blurred Boundary
Na-adjective vs. noun in Japanese is not a clean either-or choice. A lemma such as 元気 can be listed in the dictionary as both 名詞 (noun) and 形容動詞 (na-adjective) at once, and Japanese grammar lets both descriptions be true of the same word at the same time.12 The boundary blurs because な-adjectives are nominal at their core. The noun versus adjective split lives in the particles around a token, not inside the word itself.345
Overview
Where this article sits in the adjectives map
This article picks up where the Japanese Adjectives Overview: The Two Classes (い-形容詞 vs な-形容詞) and な-Adjective Conjugation in Japanese: All Tenses and Forms leave off. The two-class split (い versus な) and the eight-cell paradigm with the copula are assumed N5 background.67 This article covers the N4 boundary case: lemmas listed as both noun and な-adjective in the dictionary, and the diagnostic that tells which use a token has in a given sentence.812
The conjugation paradigm itself is not re-derived here.
The question this article answers
Learners often ask whether a word like 元気 is "really" a noun or "really" an adjective. The dictionary answer in JMdict and on Jisho is "both": 元気 is tagged Na-adjective and Noun on the same entry. So are 健康, 自由, 平和, 安全, 必要, and many other abstract-concept lemmas.12
This can be true because な-adjectives cannot carry tense or negation by themselves. They leave that work to an attached copula, exactly as nouns do.345 The part-of-speech split therefore lives in syntax (the surrounding particles), not in the word's morphology.
Why the boundary blurs in one line
Both な-adjectives and noun predicates leave tense and polarity to an attached form of the copula (だ, です, だった, でした, じゃない, ではない).359 The single syntactic slot that separates the two classes is the attributive slot, the position directly before a following noun: na-adjectives insert な, and nouns insert の.10911 Both な and の are grammaticalised attributive markers from an older pattern. な is a contracted reflex of the classical copula's attributive form (なり > naru > na, originally に + あり). の is the genitive-and-attributive particle that has linked nouns since Old Japanese.129 The fuller copula history is covered in The Japanese Copula: です, だ, である Explained.
Why な-Adjectives Behave Like Nouns
What "no internal inflection" means
A Japanese verb inflects on its own stem: 食べる shifts to 食べた and 食べない without help.513 An い-adjective does the same: 大きい shifts to 大きかった and 大きくない.10911
A な-adjective is different. The base does not change shape across the paradigm; tense and polarity live on an attached form of the copula だ.359 A noun predicate behaves identically: the noun does not change shape, and the same copula carries the same work.51014
| Class | Citation form | Past affirmative | Negative |
|---|---|---|---|
| Verb | 食べる | 食べた | 食べない |
| い-adjective | 大きい | 大きかった | 大きくない |
| な-adjective (+ COP) | 静かだ | 静かだった | 静かじゃない |
| Noun (+ COP) | 学生だ | 学生だった | 学生じゃない |
The bottom two rows follow the same paradigm, with a different lemma in front of だ. The top two do not.345 The parallel inflecting class is mapped in い-Adjective Conjugation in Japanese: All Tenses and Forms.
学生だ。10
"It is a student."
静かだ。10
"It is quiet."
学生じゃない。6
"It is not a student."
静かじゃない。6
"It is not quiet."
The morphological profile is shared: neither the noun 学生 nor the な-adjective base 静か hosts inflection internally.345
The copula carries the work in both classes
Every form of a な-adjective predicate is a bare base plus a form of the copula (静かだ, 静かだった, 静かじゃない, 静かでした).3109 Every form of a noun predicate is a bare noun plus the same form of the copula (学生だ, 学生だった, 学生じゃない, 学生でした).51014 The two paradigms are identical except for the lemma's part-of-speech label in the dictionary. This is what Nishiyama (1999) and Uehara (1998) capture by classifying な-adjectives as a sub-type of nominal predicate.34 Eleanor Jorden's term na-nominal in Japanese: The Spoken Language encodes the same insight at the pedagogical level.15
田中さんは学生でした。6
"Mr Tanaka was a student."
田中さんは元気でした。6
"Mr Tanaka was well."
部屋は静かだった。10
"The room was quiet."
Where they diverge: the attributive slot
The one syntactic slot that separates the two classes is the attributive slot, the position directly before a following noun. A な-adjective inserts な (静かな部屋); a noun inserts の (学生の部屋).10911 The な here is the modern reflex of the classical copula's attributive form なり. The の is the genitive-and-attributive particle that links one noun to another.12916 For a single use of a token, the two slots are mutually exclusive. That is what makes the slot diagnostic.917 The の particle's full range is covered in The の Particle: Possessive, Nominalizer, Attributive.
学生の部屋。10
"A student's room."
The 健康な人 versus 健康の問題 minimal pair is the cleanest worked example for this boundary: same lemma, same speaker, different particle, different relation to the following noun.8217
The 形容動詞 Classification Debate
School grammar (国語文法): the verb-like adjective
In Japanese school grammar (学校文法 or 国語文法), the framework taught in Japanese schools and used in 国語 textbooks, な-adjectives are 形容動詞, literally "adjectival verbs" or "verb-like adjectives".8189 Haga Yaichi first applied the term to this class in 『中等教科明治文典』 (1904).199
The school-grammar logic is morphological. Because the class inflects (with help from the copula), it cannot be 名詞 in the school-grammar system, where nouns do not inflect at all.8189 The 形容動詞 label captures the inflection fact, but it can obscure the nominal-syntax fact.345
Descriptive linguistics: the nominal adjective
Western descriptive linguistics calls the same class "nominal adjectives," "adjectival nouns," or "na-nominals".203515 Samuel E. Martin's A Reference Grammar of Japanese (Yale, 1975) adopts "adjectival noun" as the standard English label and is the most widely cited Western reference for the term.20
The descriptive logic is syntactic. These words attach to the copula exactly as nouns do, take case particles in their noun use, and differ from nouns only in the attributive slot.345 Nishiyama (1999) formalises this in a Distributed-Morphology analysis: な-adjectives and nouns share a phrase structure and differ only at the level of surface morphology.3 Uehara (1998) reaches the same conclusion from a cognitive-typological angle, classifying な-adjectives with nouns rather than with verbs.4 Shibatani (1990) treats な-adjectives as nominal predicates in the survey grammar.5
Both labels are right about different facts
School grammar and descriptive linguistics are not disagreeing about Japanese. They are weighting different facts about the same class.3489 形容動詞 weights inflectional behaviour, which is verb-like once the copula is attached. "Nominal adjective" weights distributional behaviour, which is noun-like at the lemma level.34818
If you read Japanese-authored materials, you need to recognise 形容動詞 as the school-grammar label.89 If you are reasoning about how the words combine in a sentence, the descriptive label is more accurate. Both should be in your toolkit.
Dual-Class Words: Lemmas That Are Both
What "dual-class" means here
A dual-class lemma is listed in the dictionary under more than one part-of-speech tag, typically 名詞 plus 形容動詞, and sometimes also の-adjective and adverb.82112 In the JMdict and EDICT tagging scheme used by Jisho, the relevant tags are n (noun), adj-na (adjectival noun / 形容動詞), adj-no (noun which may take の), and adv (adverb). A single headword can carry multiple tags on the same sense.1
This is not a small exception list. Dual-class behaviour is concentrated in the Sino-Japanese (漢語) vocabulary stratum, where many abstract-concept words function in both slots without modification.9 A large share of Japanese nouns are 漢語, and many of these abstract-concept lemmas carry both n and adj-na tags in the dictionary.9 Your job is not to memorise the list. It is to recognise the pattern and read the dictionary tags.
The N5 to N3 roster
The dual-class lemmas a JLPT N5 to N3 learner will meet in the wild fall into rough semantic domains. The JMdict tag pattern is the primary evidence.12
| Lemma | Gloss | JMdict tags | Notes |
|---|---|---|---|
| 元気 genki | energy / well-being | adj-na, n | dual-class, balanced split |
| 健康 kenkō | health / healthy | n, adj-na | dual-class, cleanest minimal pair |
| 自由 jiyū | freedom / free | n, adj-na | dual-class |
| 平和 heiwa | peace / peaceful | n, adj-na | dual-class, leans noun |
| 安全 anzen | safety / safe | n, adj-na | dual-class |
| 必要 hitsuyō | necessity / necessary | adj-na, n | dual-class |
| 普通 futsū | normal / normally | adj-no, adj-na, adv, n | leans の in attributive |
| 自然 shizen | nature / natural | n, adj-na, adv | dual-class |
| 真実 shinjitsu | truth / real | n, adj-no, adj-na, adv | triple-listed |
| 静か shizuka | quiet | adj-na (only) | adjective-only contrast point |
静か is dictionary-tagged adj-na only, not n. It is the clean contrast point showing that not every な-adjective is dual-class. The nominal "quietness" exists in the language, but only as the derived form 静けさ (built with the -さ nominaliser), not as a bare 静か.82112 真実 carries four tags (n, adj-no, adj-na, adv), and Tofugu treats it as a worked example of a lemma listed across noun, な-adjective, and の-adjective at once.217
The split between noun-leaning and adjective-leaning members varies by lemma. 元気 and 健康 sit closest to a balanced split between noun and adjective uses in general corpora. 平和 leans noun (the phrase 平和を願う and similar collocations dominate news writing). 静か is overwhelmingly adjectival.91722
Worked pair: 元気 as both
元気 is the canonical search-intent anchor for the "is it a noun or an adjective" question, and Jisho lists it as adj-na, n.2 In its noun use, it carries case particles directly: が in 元気がある, を in 元気を出す, and で in the set farewell お元気で.1082 In its adjective use, it takes the attributive な (元気な子), the predicative copula (元気だ / 元気です), and the adverbial に (元気に走る).1062
Noun use.
今日は元気がある。2
"I have energy today."
もっと元気を出してください。8
"Please cheer up."
お元気で。8
"Take care."
Adjective use.
元気に走っている。6
"Running energetically."
The same lemma 元気 carries case particles (が, を) in the noun row and attaches に, な, or だ in the adjective row. The word itself does not change. The particles do all the sorting.
Worked pair: 健康 as both
健康 is the second canonical pair because its noun-use sentences look more obviously noun-like to a beginner than 元気's do. Jisho tags 健康 as n, adj-na.2 In the noun use, it takes the genitive の (健康の問題), the goal particle に (健康に気をつける), and the purpose phrase のために (健康のために).8212 In the adjective use, it takes attributive な (健康な人), the predicative copula (健康だ / 健康です), and the adverbial に (健康に育つ).8212
Noun use.
健康の問題がある。8
"There is a problem about health."
健康のために歩きます。8
"I walk for my health."
Adjective use.
子どもは健康に育っています。6
"The child is growing up healthily."
The contrast 健康な人 ("a healthy person") versus 健康の問題 ("a problem about health") is the sharpest minimal pair for the boundary question. The lemma 健康 is identical, but the following noun is different. The particle in between switches, and the relation switches with it. な signals "the noun has this quality"; の signals "the noun is about this topic".8217
Worked pair: 自由 and 普通 in brief
自由 is tagged n, adj-na in JMdict.2 The noun use takes が ("freedom is wanted") and を ("seek freedom") directly; the adjective use takes な (自由な国) and だ (自由だ).82
自由がほしい。2
"I want freedom."
自由な国に住んでいます。8
"I live in a free country."
普通 is tagged adj-no, adj-na, adv, n in JMdict. The adj-no listing is placed ahead of adj-na, reflecting a corpus pattern in which 普通の人 is the dominant attributive form in modern Japanese, while 普通な人 is attested but read as marked in many registers.21722
普通だと思う。8
"I think it's normal."
The lean toward の in the attributive position sets up the next section. 普通 is the salient case where dictionary tagging and corpus usage do not line up, and the language is mid-drift.
The Diagnostic: の vs な and the Subject Test
Step 1: the の vs な slot test (which use right now)
The core diagnostic for any single token is the attributive slot. Put the word in front of a noun and ask which particle the language wants.1091117
The "both grammatical" branch is the boundary case. 健康な問題 parses as "a healthy problem," which is semantically odd. 健康の問題 means "a problem about health," which is sensible.17 When both readings exist with different following nouns, the lemma is dual-class. The speaker is making a real semantic choice with the particle: な attributes a quality to the noun, while の links the noun to a topic or domain.8917 Tofugu frames the same diagnostic as "な is used when you want to describe a noun, and の is used when you want to label a noun." In that account, な is compatible with degree adverbs such as とても and すごく, while の carries objective categorisation.17
Step 2: the subject and object test (is this lemma dual-class)
Step 1 sorts a single token in a single sentence. Step 2 sorts the lemma itself. The lemma-level diagnostic asks whether the bare word can serve as the subject of a sentence with が, or the direct object of a verb with を, without any な or の added.1082
If yes, the lemma is dual-class. 元気がある (元気 + が + ある) is grammatical, so 元気 has a real noun use.82 健康を保つ (健康 + を + 保つ) is grammatical, so 健康 has a real noun use.82 静かがある and 静かを保つ are not grammatical. To express "quietness" as a noun, the language uses the derived nominal 静けさ (with the -さ nominaliser).8219 静か is therefore not dual-class.
健康を保つ。8
"To maintain health."
The last example uses 静けさ, not 静か, because a bare 静か cannot fill the subject slot. The subject-or-object test lets you predict the dictionary listing without looking the word up: if the bare word survives as a subject or object, it is dual-class.82
When the diagnostic disagrees with the dictionary
A small number of lemmas drift between classes by register and era; 普通 is the salient case at N5 to N3.8217 JMdict lists 普通 as adj-no, adj-na, adv, n, with adj-no placed ahead of adj-na. This reflects a corpus pattern in which the の-attributive form 普通の人 is the dominant choice and the な-attributive form 普通な人 is attested but marked in many registers.21722 国語 reference dictionaries differ on whether to fold 普通 into 形容動詞 or to give the の-attributive use its own sub-entry, and this is a known boundary case rather than a clean rule.821
For lemmas like 普通, where the dictionary records adj-na but natural attributive usage prefers の, follow the corpus pattern in production. Treat any な-versus-の choice that sounds odd in listening as a clue that the word is mid-drift.1722 The parallel question of words tagged adj-no that act like adjectives but only take の attributively (病気の人, 緑の葉) is the open territory adjacent to this article.
普通な日だった。17
"It was an ordinary day."
普通だ。8
"It's normal."
The の form is the corpus-dominant choice; the な form is attested but reads as marked.
Good to know
Reading multi-tag dictionary entries as records, not hedges
When you look up 元気 in a J-E dictionary and see "(n., adj-na) health, vigour, energy," the multiple tags are accurate, because the lemma actually has both uses.12 The same applies for 健康 (n., adj-na), 自由 (n., adj-na, adj-no), and 真実 (n., adj-no, adj-na, adv).2 JMdict and Jisho (which uses JMdict as its data source) record the attested distributional patterns of each lemma. Read both tags as accurate, then read the surrounding particles to see which one is active in a given sentence. Picking "the real" part of speech is the wrong move.12
Forcing な onto a noun-only lemma
A noun that is not dual-class never takes な in the attributive slot. 学生 is a noun, not a dual-class lemma. の fills its attributive slot, and 学生な部屋 is ungrammatical for "a student's room."109
学生の部屋。10
"A student's room."
な is reserved for the な-adjective class and for dual-class lemmas in their adjective use.109 The same diagnostic in reverse catches the matching mistake: adding な to a plain noun.
Treating 静か as if it were dual-class
静か is the clean counterexample. It is dictionary-tagged adj-na only. The nominal "quietness" is the derived form 静けさ, built with the -さ nominaliser.8212 The subject-test diagnostic catches this: a bare 静か cannot serve as a subject with が. The pair 静か versus 静けさ is the cleanest way to see what "dual-class" buys a lemma. 元気 and 健康 do not need a derived nominal because the bare lemma already covers the noun slot, while 静か does need one.
Why な and の split the attributive slot historically
In classical Japanese, the copula had two attributive forms in this functional space: なり (used with the adjectival-noun-like lemmas, derived from に + あり) and の (the genitive particle, used to link one noun to another).12916 Modern な is a contracted reflex of なり (なり > naru > na). Modern の is the same genitive-and-attributive particle, now broader in scope. The modern split between な-attributive and の-attributive is therefore a fossilised reflex of an older two-form distribution. That is why the language can still encode "is this an adjectival modification or a noun-labelling modification" cleanly in one slot. The fuller copula history is in The Japanese Copula: です, だ, である Explained.
The "if the copula carries it, it's nominal" rule
The one-line mental hook is this: any word whose tense and polarity live on an attached copula (だ, です, だった, でした, じゃない) is nominal at the core.3510 This is true for nouns and for な-adjectives by definition. It is false for verbs and for い-adjectives, which inflect on themselves. Internalising the rule turns the question "is X a noun or an adjective?" into a syntactic question about the slot the token is filling: case particles (が, を) versus attributive な versus adverbial に. The same lemma can appear in any of those slots. The slot, not the lemma, is what the question is really asking about.
Why "both, simultaneously" is the honest answer
The question "is 元気 a noun or an adjective?" assumes that part-of-speech labels are exclusive, but in Japanese they are not.812 A single lemma can carry two part-of-speech entries at once. The speaker chooses which one is active each time the word is used by choosing the surrounding particles: が or を for the noun use, な or だ or に for the adjective use. This is structurally normal in Japanese in a way that has no direct English analogue. Treating it as normal, rather than as an exception to memorise, is the actual N4-level upgrade. The dictionary tags, the の vs な test, and the subject test are all routes to the same syntactic answer.
See also
- Japanese Adjectives Overview: The Two Classes (い-形容詞 vs な-形容詞)
- な-Adjective Conjugation in Japanese: All Tenses and Forms
- Attributive vs. Predicative Adjectives in Japanese: Why 静か Needs な But 大きい Does Not
- The Japanese Copula: です, だ, である Explained
- The の Particle: Possessive, Nominalizer, Attributive
- Parts of Speech in Japanese: The 10 Classes (品詞)