1. Introduction

To regard the world as a net of cause and effect is an integral part of human thinking. It is therefore not surprising that many languages exhibit an explicit causative verb inflection, as Finnish -ttaa, Turkish -dir (with the appropriate vowel harmony), Japanese -(s)ase, Akkadian ša‑. Hence for example Finnish syö-dä eat, syö-ttää feed; Turkish ye-mek eat, ye-dir-mek feed; Japanese tabe-ru eat, tabe-sase-ru feed; Akkaddian akālu(m) eat, š-akālu(m) feed. Such an inflection is notably absent from most of Indo-European, where periphrastic constructions are used instead, as English make eat, with the exceptions of Lithuanian -in‑ and Persian -ân‑.

Another manifestation of this principle is the causative alternation (a much-studied topic, see e.g. Levin and Rappoport (1995) chapter 3, Pinker (2007) chapter 2 and the references therein), an opposition between a verb that typically denotes a change of state and another one that causes this state. There are language-specific restrictions on which verbs can participate in such an alternation. In English, for example, the verb “split” can be used transitively in a causative sense, as in “Eru split the ground”, but also intransitively, as in “The ground split”. On the other hand, the verb “cut” can only be used in a causative sense, “He cut the rope” and not **“The rope cut”; while the verb “bloom” can only be used intransitively, “The flowers bloomed” and not **“Spring bloomed the flowers”.

It is easy to see why this happens: The cutting event is difficult to dissociate from the agent doing the cutting, while the blooming event is perceived as being internally triggered, with only indirect influence from outside. On the other hand, verbs like “split” or “break” are in the middle: Such changes of state can either occur on their own as well as due to an agent that is unknown or chosen to be left out (“The ground split”); or otherwise with an explicit agent given (“Eru split the ground”). Observing that such contrasts appear across the world’s languages, one can suspect that the underlying cognition is universal. However, the manifestation is different in different grammatical environments, and may be more apparent due to the distinct morphology of the participating verbs. For example, Japanese sak-u split is transitive, while sak-e-ru is intransitive and can be more explicitly paraphrased as get split, split itself, undergo splitting.

As mentioned, the restrictions of the alternation are language-specific. For example, in English, the verbs “laugh” and “cry” do not alternate, while they do in other languages. In Japanese, transitive tor-u take alternates with intransitive tor-e-ru vanish, disappear (of illness, fatigue etc.). In order to compare alternation patterns between different languages, a further semantic restriction is commonly made in that the non-causative partner of the pair should denote a change of state (break, melt, open etc.) or a so-called “going-on”, an activity where the agent is left out (boil, spin, roll) and it can be conceptualized as occurring spontaneously. Following Haspelmath, such a verb is called “inchoative” (Haspelmath 1993). The causative counterpart1 presents the same situation with an agent. The following morphological patterns can be identified cross-linguistically, taken from Haspelmath (ibid.), though the work goes back to the Leningrad school of typology (Холодович 1969):

1. Causative alternation (C): The causative partner is derived, as Indonesian patah break (intr.), me-matah-kan break (tr.)
2. Anticausative alternation (A): The inchoative partner is derived, as Turkish kır-mak break (tr.), kır-ıl-mak (intr.)
3. Equipollent alternation (E): Both partners are derived from the same stem, as Japanese kow-are-ru break (intr.), kow-as-u break (tr.)
4. Labile alternation (L): There is no morphological difference between the partners, as English “break” = become broken or cause to break; or German zerbrechen
5. Suppletive alternation (S): Both partners are derived from different roots, as Russian гореть (goretʹ) burn (intr.) and жечь (žečʹ) burn (tr.)

From Japanese, an interesting thing can be learned here: Consider the pair ak-u open (intr.) and ak-e-ru open (tr.). One might be tempted to deduce that the morpheme -e‑ is a causative marker, but this is invalidated by the already mentioned sak-u split (tr.), sak-e-ru split (intr.), where it has the completely opposite role. We have to conclude that no meaning can be assigned to -e‑ on its own, it rather flips the causativity of whichever verb of the pair is (by convention) the basic one. The causative/inchoative pairs are thus in a way inseparable: they can only be understood in opposition to each other and not independently.

2. Transitive/intransitive pairs in late Quenya

Alternating transitive/intransitive verb pairs emerge in Quenya throughout Tolkien’s writings. Let us focus first at the later years (especially Words, Phrases and Passages and the Quenya Verb Structure). In the following, I present the attested examples grouped according to the alternation patterns introduced above. Note that I always supply a hyphen when citing verbal stems for better readability (which Tolkien also did frequently, but not consistently).

It will be seen that the verbs participating in the alternation in Quenya go beyond the causative/inchoative subset defined above. Quite often, one of the partners does not express a change of state at all, but rather the state itself. For this reason, I will prefer to use the terms “transitive” and “intransitive” whenever referring to the specific sets of alternating Quenya verbs (which are grammatical rather than semantic terms), but will continue to use “causative” and “inchoative” when talking about the particular semantics. Terms like “equipollent” and “labile” refer to the morphology and can be kept. In any case, a slight blurring of the terminology seems unavoidable.

3. Analysis in terms of a hierarchy

Simply looking at the above examples synchronically, without delving into the historical development (which comes in the next section), one observes that the ending ‑ya defies a simple translation (similar to the -e‑ morpheme in Japanese discussed above): While it takes the role of an anticausative marker in words like virya‑, it is perfectly capable of taking the role of a causative marker in komya‑ as well. We have thus to look at the alternation pattern: One observes that in the former case, the contrast is formed with ta-verbs (vista‑), but in the latter case with root verbs (okom‑). I therefore propose the analysis that there is an underlying “causativity hierarchy” in Quenya which can be represented in the following fashion:

• [more causative] ‑ta > ‑ya > root verb (> explicit anticausative markers) [less causative]

This means the following: The suffix ‑ta occupies the highest position and will mark the causative in any pair. The ending ‑ya is in the middle and can mark the causative in contrast with root verbs, but will mark the anticausative in contrast with ta-verbs. Finally, root verbs are in the bottom position and can only be anticausative when contrasted with a ‑ta/‑ya-verb. If we are dealing with labile alternation, any verb type of the three seems to have the potential to be both causative and anticausative (although there appear to be no examples for root verbs). Apart from that, the endings ‑ta and ‑ya can form other types of verbs as well, but here the focus is on alternating pairs.

The mention of “explicit anticausative markers” in the hierarchy refers to the inceptives and cases as the reflexive quere immo *turn (oneself). Having the above mechanism with the endings ‑ta and ‑ya in place, it seems redunant to have a widespread reflexive inflexion to mark inchoatives as is the case across Romance, Germanic, Slavic and Greek. Notably, the root verb quer‑ turn is already causative in nature (it has kendele or kanwa face as direct object in other versions of the Ambidexter’s Sentence). According to the hierarchy model, any derivation with ‑ta, ‑ ya would render this only even more causative, perhaps *querta‑ make turn, which may be why the reflexive is employed here. We can therefore speculate that the reflexive in Quenya is limited to actual cases of the agent and patient being the same (as wash oneself etc.) and is not used in an inchoative context where the agent is simply left out and the situation is presented as occurring spontaneously.

The hierarchy model fits all of the examples above except for lerta‑/lerya‑, which essentially seem to be be free and make free respectively. However, they are not listed beside each other by Tolkien, so it may be a case of changing his mind. Also, lerta‑ is rather used as an auxiliary verb lerta[n] quete I can speak because I am free to do so, there being no obstacle of promise, secrecy, duty, suggesting that there is not really an alternation between the two.

The model may also explain why Q. koita‑ live, be alive (PE22:103), live, have life (PE22:125) describing a state easily switches to causative koitā́‑ bring to life (PE22:136, with only the Common Eldarin form given). In the former case, it appears on its own, but in the latter case in contrast with et-koiru‑/lu‑ come to life, making the ta-verb assume a higher position in the hierarchy.

4. History of the endings

The Quenya Verb Structure (PE22) offers a historical perspective on how the above system came about. The notes are often difficult to follow, but the main idea is the following: The Common Eldarin endings ‑tā and ‑jā were both causative in nature3. They have to be distinguished from short ‑ta, ‑ ja (often with a breve: ă) which Tolkien calls “formative” (PE22:134-135). The formative endings are required to form proper verbs from the corresponding roots without altering the meaning, so for example the root SIR‑ flow yields the verb sirya‑ flow. When applied to adverbial roots like OR/RO‑ up, they form inchoatives, in this case orta‑, orya‑ rise (in contrast with causative ortā‑, oryā‑ raise)4.

However, given the alternation patterns discussed in the previous section, this system has to restructure or restrict itself at some point, so that in a pair like vista‑/virya‑ the former verb hails from a causative *wistā‑, while the latter stems from a formative *wisjă‑. In a pair like komya‑/okom‑, however, the former verb probably stems from a causative *komjā‑. The crucial thing is that this seems to happen in accordance with the hierarchy rule established above, so that on the surface of later Quenya, we always see ya-verbs as more causative than root verbs and ta-verbs as more causative than both ya-verbs and root verbs. This process is not described in detail except for a brief note that to avoid the confusion with the causatives ‑ya was preferred for intransitives: so oryane, rose, ortane, raised (PE22:115); and a similar note in PE22:157.

5. Inflexion or derivation?

As mentioned in the introduction, another manifestation of causativity in many languages is the causative inflexion. In principle, we can have the following three layers of morphology: Inflexions are productive (i.e. they can be predictively formed from all eligible words) and do not change the part of speech. Derivations are only productive to a certain extent, not always easily predictive, and often change the part of speech. For example, the English derivative suffix -le has a frequentative sense in a number of words crackle repeatedly produce cracking sounds, but is far from being productive, so that there is for example no **clickle repeatedly produce clicking sounds. The final layer is lexical, like “die” and “kill”, which are semantically related, have a common root in some languages (like Turkish öl-mek die, öl-dür-mek kill, lit. make dead) and yet are completely different in English.

While the causative in languages like Japanese fulfills the criteria of an inflexion, the Quenya causatives ‑tā, ‑yā have rather to be seen as derivational for the following reasons:

1. Tolkien only gives a limited number of examples and never lists the suffixes in any of the numerous verbal conjugation tables, suggesting that they are of limited productivity.
2. The same suffixes are frequently used in deriving verbs from adjectives, thereby changing the part of speech: niñkwitā‑ make pale, white, karan-yā‑ make red, redden (PE22:114).

However, in early Qenya Tolkien mentions the verbal prefix ka‑ (from KAHA cause PE12:43) that might be a genuine causative inflexion on top of alternating pairs (which are discussed further below). Two examples are given : kamanta‑ to make eat, give to eat and kasóro‑ make sit, set. Such an inflexion is not attested for later Quenya.

Curiously, the example kamanta‑ has an additional nuance give to eat. Indeed, causative inflexions across natural languages tend to have such additional nuances as permissive let do, allow to do, assistive help to do and so on (Недялков & Сильницкий 1969, pp. 28-31; Payne 1997, p. 178). The precise translation is often dependent on the context.

6. Transitive/intransitive pairs in early Qenya

Tolkien’s earliest vocabulary compilations, the Qenya Dictionary (PE12) and the Qenya Word-lists (PE16:129-145) already had its share of alternating verb pairs. The dominant pattern is a causative alternation between transitive -ta5 and a root verb:

• KALA: kala‑ shine, kalta‑ set light to, kindle (PE12:44)
• KAPA: *kapa‑ jump, kapta‑ startle (PE12:45) [the former is from kapin I jump, the latter is apparently literally make jump]
• KELE, KELU‑: kelu‑ to flow, kelta‑ draw water (PE12:46)
• *KṆKṆ: †kinka‑ to hang (intr.), kinkata‑ hang (tr.) (PE12:47)
• LESE: lese‑ come together, gather (intr.), leste‑ collect (tr.) (PE12:53)
• NAẎA: naẏa‑ impers. it grieves, naitya‑ damage, hurt (PE12:65) [with *-kə-ya > ‑ tya]
• NŌ‑: nō‑ be born, become, nosta‑ give birth to, cause (PE12:66)
• ORO: oro‑ rise, orto‑ raise (PE12:70)
• QOŘO, QOSO: qoro‑ choke, suffocate, esp. = drown (intr.), qosta‑ choke, drown (tr.) (PE12:78)
• RUKU: ruku‑ to steam, smoke, reek, rukta‑ belch smoke, spout smoke, cause to smoke; smoke a pipe (PE12:80)
• TALA: tala‑ carry, bring; weigh (intr.), talta‑ to lade, burden, load, charge; enjoin; oppress, weigh down (PE12:88)
• ULU: ulu‑ pour, gush (intr.), ulto‑ pour (tr.) (PE12:97)
• VASA: *vasa-, vasta‑ to rush (tr. & intr.) (of both noise and speed) (PE12:100) [These are given in what must be the 1st singular inflection vasin, vastan. Taken literally the gloss would mean that former verb is intransitive and the latter transitive, but given the vast number of examples where it is the reverse, the labels just probably do not match the order of the verbs in this case.]
• ere goes, esta‑ send (PE16:133) [deleted; the root appears to be more or less identical to ERE‑, ESE‑ out in PE12:36]
• hyal‑ ring, resound, hyalta‑ to strike, make ring (PE16:144)
• lev‑ move (intr.), lehta‑ move (tr.) (PE16:132)

An interesting use of the ending ‑ta is seen in the pair tyasa‑ test, try, pick, choose, tyasta‑ put to the test (PE12:49, TYASA). Both are transitive and a causative variant of the former would mean make someone test/try/pick/choose. Instead, put to the test may be seen to rank higher in the category of affectedness. Thus causativity may be correlated with affectedness, compare also the polysemy of talta‑ between lade, burden, load, charge and the more affective oppress, weigh down.

In some cases, the transitive variant is formed by the ending -u and the intransitive variant denotes a state:

• KALA: kala‑ shine, kalu‑ illuminate, light up (PE12:44)
• KAMA: kama‑ to lie down, c. loc. to endure, suffer, kamu‑ to lay down, bend down, reduce (PE12:44)

Causative alternation using -ya is much more rare in early Qenya:

• PUŘU: puru‑ burn, purya‑ set fire to (PE12:75)
• tul‑ come, tulya‑ bring (PE16:133)

Equipollent alternation appears in a couple of examples:

• KAYA: kaima‑ lie quiet, kaita‑ to place (PE12:46) [apparently lit. make lie]
• LINI‑: linya‑ run or flow smoothly, linta‑ soothe (PE12:53) [Whether the latter is a causative variant of the former does not seem entirely certain.]
• MINI: minda‑ to diminish, fade, lessen, vanish, minu‑ make less, decrease, spoil, alter for worse (PE12:61)
• SAẎA: saita‑ (saya-) impers. I am hungry, saitya‑ to starve (tr.) impers. = I starve (PE12:82) [with *-kə-ya > ‑tya]
• TUŘU, TUSO: turya‑ catch fire (PE12:96), tunda‑ kindle [cf. †turu‑ below]
• USU, UŘU: urya‑ burn (intr.), usta‑ burn (tr.) (PE12:98)

Of these, kaima‑ is quite unusual, since ‑ma is normally not a derivative ending for verbs. Perhaps we are dealing with a denominal verb here from the noun kaima bed. Note that kaima‑ was changed from kaito‑ at the same time as kaitosambe bedroom was changed to kaimasambe.

Similarly, minda‑ is unusual as ‑da is normally not a verbal derivation suffix. Perhaps it is an analogical formation to roots containing Ř, where ‑nd‑ results from a nasal infix. In this case we would be dealing with an originally causative alternation belonging to the same group as kala‑/kalu‑ and kama‑/kamu‑ above.

A few examples show labile alternation:

• OLO: olto‑ increase, multiply (tr. and intr.) (PE12:69) [note however tr. olta‑ magnify, extol, praise right beside it]
• SAẎA: saitya‑ to starve (tr.) impers. = I starve (PE12:82) [with *‑kə-ya > ‑tya]
• TṆTṆ: tanta‑ bounce, bound, rebound, tanta‑ to dance (tr.), dandle, wave (PE12:94)
• TULU: tulu‑ (1) bring, carry, fetch (2) intr. move, come (3) produce, bear fruit (PE12:95)

There is an unusual suppletive example which seems to result from a phonetic variation between q and p:

• QIŘI: qiri‑ qirna‑ stir (make spin), PIRI: piri‑ spin, turn (PE12:74,77)

Finally, one example can be interpreted as anticausative alternation, where the anticausative partner is derived with ‑ya. However, apart from †turu‑, there is also the variant tunda‑ (see above):

• TUŘU, TUSO: †turu‑ kindle, turya‑ catch fire (PE12:96)

7. Semantics of the participating verbs

It was mentioned that the set of verbs that can participate in alternations in Quenya is larger than the inchoative group. This section attempts to group them according to their semantics with the help of the categories established by Levin and Rappoport (1995). For better readability, only the basic translated senses will be listed in the following, see above for the fully glossed Quenya examples.

Various agentive verbs of motion, manner of motion and crossing often allow for a causative counterpart in Quenya. Since these involve an agent performing the action and not a spontaneous movement, they rarely participate in causative alternations across languages. In English some may have a very limited transitive use, as in “jump a horse over the fence”. Note also that verbs of crossing as cross, reach, enter naturally have the destination as their direct object.

• go / make go, send
• come / make come, fetch
• reach / make reach, send
• jump / make jump, scatter (or startle)
• run / make run (ride)

Verbs of (manner of) motion that are not necessarily agentive, such as move, spin, roll, rush, do participate in alternations. In these cases the movement can be conceptualized as happening on its own once the external cause is gone. In our Quenya examples, this applies to the following:

• turn (intr.) / turn (tr.)
• spin (intr.) / make spin, stir
• move (intr.) / move (tr.)
• rush (intr.) / rush (tr.)
• bounce / make bounce, dandle

Another common group are the verbs of emission (of light, sound, substance). These are triggered internally rather than by an external agent and hence generally do not participate in alternations across languages. However, an external agent may manipulate the emitter which gives rise to a causative sense. Again, this is quite limited in English, an example would be “ring a bell”.

• shine / cause to shine
• glint / cause to glint
• ring / cause to ring
• smoke / cause to smoke
• gush, pour (intr.) / pour (tr.)
• start flowing / cause to flow

Perhaps related to the above group are the verbs of appearance. Note that the categories are not mutually exclusive, so that come and gush, pour, flow may be counted here as well:

• awaken (intr.) / awaken (tr.)
• come to life / bring to life
• be born / beget
• come / make come
• gush, pour (intr.) / pour (tr.)
• start flowing / cause to flow

We have some examples of verbs that denote mental states or physiological states. Once more, these arise internally and do not normally alternate. However, certain mental states may have an obvious external trigger which gives rise to a causative variant.

• be excited / excite
• feel terror / terrify
• grieve / hurt
• starve (intr.) / starve (tr.)

Finally there are a few verbs that refer to spatial configuration. Certain verbs of this kind allow a transitive use in English, where the configuration is imposed by an external agent (such as hang or lean).

• hang (intr.) / hang (tr.)
• lie / place
• be directed / direct

8. Comparison with natural languages

An interesting typological comparison with regard to causative/inchoative pairs was carried out by Martin Haspelmath (1993). He first notes that there are verbs where the semantics strongly require an agent (such as wash for example) and so the primary verb of the pair will practically always be causative. For other verbs (such as dry), the situation can occur spontaneously often enough, so that they can exhibit very different alternation types in different languages. He put together a list of 31 verbs which fall into this category:

1. wake up 2. break 3. burn 4. die/kill 5. open 6. close 7. begin 8. learn/teach 9. gather 10. spread 11. sink 12. change 13. melt 14. destroy/be destroyed 15. lose/get lost 16. develop 17. connect 18. boil 19. rock 20. go out/put out 21. rise/raise 22. finish 23. turn 24. roll 25. freeze 26. dissolve 27. fill 28. improve 29. dry 30. split 31. stop

He than compared 21 languages from different families regarding which alternation patterns they use for these 31 verbs. Unfortunately, due to gaps in attestations, we can only make a partial list in Quenya. Substituting lower/descend for sink and using the abbreviations (E)quipollent, (C)ausative, (A)nticausative, (L)abile and (S)uppletive, we get the following:

• 1. (E) kuinu‑ / kuita‑
• 3. (E) urya‑ / usta‑ or (C) puru‑ / purya‑ [early Qenya only]
• 4. (S) qual‑ (PE22:152) / nak‑, nahta‑ (PE22:112,164 etc.)
• 6. (E) holya‑ / holta‑
• 9. (C) okom‑ / komya‑ or (L) hosta‑
• [11. (E) núya‑ / núta‑ etc.: these are lower/descend or sink, set (of Sun and Moon) rather than sink in the sense of submerge]
• 12. (E) virya‑ / vista‑
• 21. (E) orya‑ / orta‑ or (L) orta‑, orya‑
• 22. (C) tel‑ / telya‑
• 23. (A) queresse, quere immo / quer‑
• 31. (L) pusta‑ [Etymologies only]

Thus, if we count 3, 9, 21 as half for each of the shared categories, we arrive at (E): 4-5, (C): 2, (L): 2, (A): 1, (S): 1. And if we count differentiated past tenses as equipollent rather than labile, it will bump up the (E) count to 6-7 and put (L) to zero. No differentiated past tenses are attested for pusta‑, hosta‑, but their existence seems very likely judging by other similar examples.

Furthermore, we have the following partial attestations: 27. quanta‑ fill (PE17:68), 30. sahta‑ split as marked causatives, so that for these cases, only (E), (C), (L) are possible (apart from (S) which is always possible); 27. quat‑ fill (XI:392) and 2. rak‑ break are unmarked causatives (the participle rákina broken implies that the verb is transitive), so that only (A), (L) are possible; finally 10. palu-, palya‑ to open wide, spread, expand, extend are probably marked inchoatives (due to the inchoative tendency of ‑u discussed above), so that only (A), (E), (L) are possible here.

So where does Quenya stand typologically when compared to natural languages? Instead of reproducing the detailed statistics of all the 21 languages, let me attempt to lump them together into similar groups. Some languages strongly prefer a single type of alternation (with a count of more than 20 of the possible 31):

• English: L
• Romanian, Russian, Hebrew: A
• Japanese: E
• Mongolian: C

Other languages prefer two types of alternation with more or less equal weight (something like 12-17 for each):

• Indonesian: E ≈ L
• Greek, German: A ≈ L
• Finnish: A ≈ C

Yet other languages have a dominant alternation (about 17-20) and a subdominant one (7-9):

• Turkish: C > A
• Armenian, Hebrew, Arabic: A > C
• French: A > L
• Georgian: E > A

Finally, there are languages that are spread between three alternations (about 7-12 each):

• Hungarian, Swahili, Hindi-Urdu: A ≈ C ≈ E

If we are allowed to extrapolate from our incomplete sample for Quenya, it looks like it prefers equipollent alternation followed by the causative or labile one, in suggestive equation form:

• Quenya: E, E > C or E > L

Thus, depending on the exact ratio between E and C, the best matches are in fact Japanese (high ratio: E=20.5, C=5.5) and Indonesian (almost equal ratio: E=17, C=14). Georgian has E=15.5 and C=4.5, but is not a good match because of its high count of A=9 which seems uncommon in Quenya. There is actually no good match in Haspelmath’s list which would be of type E > L (dominant E and subdominant L), so that if this were to be the case, Quenya would be quite distinct from all the 21 languages in his sample. Remember that this only concerns the list of selected verbs which can be equally spontaneous or agent-triggered and is not a comparison of the total inventory (which would be more difficult).

In any case, Quenya stands separate from the Indo-European languages which seem have a strong tendency towards A (except for English which is quite unique in its reliance on L, perhaps due to its limited morphology), while their E count is quite low (for Greek, Romanian and French it is actually zero).

9. Conclusion

The concept of morphemes has proven to be an effective way of analyzing and understanding grammar. However, sometimes there are slightly different principles at work. One such well-known principle is the animacy hierarchy (Payne 1997, p. 150). It is universally occurring, but manifests itself differently in various languages. In this work I have argued that Quenya exhibits a language-specific causativity hierarchy that accounts for the distribution of endings in transitive/intransitive (or semantically causative/inchoative) verb pairs. For example, the suffix ‑ya assumes an anticausative role when contrasted with ‑ta and a causative role when contrasted with a root verb.

In the second part a comparison to natural languages was attempted. Many studies of both Tolkien’s languages and his mythology seem to pay a great deal of attention towards finding his real-world inspirations and influences. However, an over-reliance on such analyses might lead to overlooking Tolkien’s own creative glossopoeic contributions. In terms of causative semantics, at least, it does not appear that the usual real-world suspects like Finnish, Greek, Latin or Germanic are of much help in elucidating Tolkien’s choices in Quenya. The preference for equipollent alternation is exotic from the point of view of Indo-European (and more generally European) and is best matched by such distant languages as Japanese or Indonesian, which can be ruled out as direct influences on Tolkien.