lemÌc. Lemizh grammar and dictionary

Unit 16. Dependent clauses: introduced with pronouns or adverbs

Ylva-li and I have a special language, which no one understands but the two of us.

(Astrid Lindgren. Allrakäraste Syster)

Pronouns and adverbs vs. conjunctions

Now we move on to finite clauses introduced with a pronoun (e.g. ‘who, which, that’) or an adverb (e.g. ‘where, why’). To get the first naming issues right out of the way: from the Lemizh point of view, the difference between subordinating pronouns and adverbs is just a matter of plot case vs. causal/temporal/spatial case. Most of these clauses also go by the name of relative clauses, but we will also treat indirect questions in this unit.

Some of the English (and other Indo-European) pronouns and adverbs introducing dependent clauses look like subordinating conjunctions but are used quite differently. Compare:

I know that the child is hiding beneath the rose bush.I know the child that is hiding beneath the rose bush.
The entrance vanished when the roses had faded.The day when the roses had faded, the entrance was gone.
I’m able to think wherever the flowers sing.I’m thinking of wherever the flowers sing.

As you see, clauses introduced with pronouns and adverbs usually occur in the role of attributes (or adverbs, as we will see shortly). This should remind you of the attributive and adverbial adjectives and participles in unit 5. However, we will also meet a group of counterexamples, the headless relative clauses.

Relative clauses

Attributive and adverbial clauses

Just as conjunctional clauses are the finite counterparts of infinitive and gerund clauses – being expanded objects –, we are now dealing with the finite counterparts of participial clauses – so we use brackets and coordinations for translation. Different relative pronouns (English, not Lemizh ones) and adverbs correspond to different inner cases: pronouns to plot cases and adverbs to others. Recall that the look-alike conjunctions correspond to different outer cases.

hrá oRwxÌfe mesÌy prilneytfÌaR.The midnight-born (-‘mothered’) poodle is yelping.The poodle that was born at midnight is yelping.
yelp-fact1 poodle-acc-nom2a mother-acc-acc3 midnight-acc-temp4.
Rájg htrybÌe Recý afkÌar drulygwrÌjde.My sister, a queen, lives beneath the rose bush.My sister, who is a queen, lives beneath the rose bush.
live-fact1 sibling-acc-female-acc-nom2a monarch-nom-acc3 up-fact-opposition-acc-loc2 bush-acc-rose-acc-nom3.
jàx nÌwi ganáry lÌbe.We went to the valley, the location of the singing of flowers.We went to the valley where the flowers sing.
move-fact1 valley-acc-dat2 sing-loc-acc3 flower-acc-nom4a.

Not surprisingly, adverbial clauses (clauses in the role of adverbs; not necessarily introduced with adverbs) are usually factive (or consecutive) brackets. Again, the pronoun or adverb is reflected by the inner case.

hrá oRwxyfè rahkÌa.The poodle is yelping, which I don’t like.
yelp-fact1 poodle-acc-nom2a like-fact-opposition-acc-fact2.
hrá oRwxyfè Ólva zeè vèi.(compare the corresponding conjunctional clause)The poodle is yelping, wherefore I’ll feed it.
yelp-fact1 poodle-acc-nom2a eat-psu-fact2 PIn−3-nom-nom3a PIn−2-nom-dat3.
fìlt yhwè tÌpil cOÌci.The horse is fast, which/as people assume.
fast-cons1 horse-acc-nom2 assume-acc-cons2 human-acc-dat3.

In this light, we can translate the inverted gerund ‘I didn’t see Amélie watching TV’ into English as the more accurate but less elegant ‘Amélie was watching TV, which I didn’t see’.

The English formal ambiguity between conjunctions and adverbs sometimes translates into an ambiguity between outer and inner case in Lemizh.

zdàs dmatmàqkar keltÌje.(conjunctional)He sat down where he could see the pendulum.
seat-fact1 see-fact-opportunity-fact-loc2 pendulum-acc-nom3.
zdàs dmatmàrqka keltÌje.(‘relative adverbial’: factive bracket)
seat-fact1 see-fact-opportunity-loc-fact2 pendulum-acc-nom3.

Restrictive relative clauses

If an attributive relative clause is not separated from the main clause by a comma, it is considered restrictive in English. Restrictive clauses can also be introduced with the pronoun ‘that’ (as opposed to ‘which’): compare ‘The poodles that/which were born at midnight are yelping’ (restrictive: there are other poodles, but I’m referring to the ones born at midnight) with ‘The poodles, which were born at midnight, are yelping’ (non-restrictive: all of them were born at midnight; and this fact is additional information about them). A restrictive clause sometimes has to be translated with a partitive bracket, as you know – becoming the bracket’s predicate.

oRwxÌf tÌy. hrá mesÌe qyÌn prilneytfÌaR.The midnight-borns from the set of poodles are yelping.There are these poodles. The ones that were born at midnight are yelping.
poodle-acc1 this-acc-acc2. yelp-fact1 mother-acc-nom2a PIIn−2-acc-partacc3 midnight-acc-temp3.
Often a partitive coordination can achieve a better word order:
oRwxÌf tÌy. hrá fyèn mesÌe prilneytfÌaR.There are these poodles. The ones that were born at midnight are yelping.
poodle-acc1 this-acc-acc2. yelp-fact1 PIIn−1-acc-partnom2a mother-acc-nom2 midnight-acc-temp3.

Non-restrictive relative clauses

A non-restrictive relative clause in English is one with commas. It can normally be translated with a bracket just like other relative clauses. However, to explicitly state that the contained information is additional and does not serve to identify the referent (here, the instance of ‘poodle’), it has to be moved into a separate sentence, which can then be connected with ‘and’ or weakly linked.

hrá oRwxÌfe mesÌy prilneytfÌaR.bracket: The midnight-born poodle is yelping.The poodle, which was born at midnight, is yelping.
yelp-fact1 poodle-acc-nom2a mother-acc-acc3 midnight-acc-temp4.
hrá oRwxÌfe. mesìl feÌ prilneytfÌaR.two sentences: The poodle is yelping. It was born at midnight.
yelp-fact1 poodle-acc-nom2a. mother-cons1 PIIn−1-nom-acc2 midnight-acc-temp2.
hrána oRwxyfé mesàna feÌ prilneytfÌaR.‘and’: The poodle is yelping and it was born at midnight.
do-fact1 yelp-partfact-fact2 poodle-acc-nom3a mother-partfact-fact2 PIIn−1-nom-acc3 midnight-acc-temp3.
hrá oRwxyfè mesìlnyl fyÌ prilneytfÌaR.weak linking: The poodle is yelping; it was born at midnight.
yelp-fact1 poodle-acc-nom2a mother-partcons-ctx2 PIIn−1-acc-acc3 midnight-acc-temp3.

The first translation is appropriate whenever the poodle can be clearly identified without the relative clause, so that it obviously provides additional information. The other translations make explicit that the midnight-born-ness is not needed to pin down the instance of ‘poodle’.

Some English non-restrictive clauses serve as explanations or reasons for the main clause and so are called explanatory relative clauses. They are translated as conjunctional clauses, typically in causal cases, like the circumstantial clauses (including the caveat about conjunctional clauses in negated sentences). Translating them as relative clauses would lose the cause–consequence relation in Lemizh.

nà smàjy yxì djáy viì qkrÌdjy, wàxel elefyÌ cnÌi.The man, because he was speaking to the child about elephants, forgot to buy the artichokes.The man, who was speaking to the child about elephants, forgot to buy the artichokes.
not-fact1 remember-fact-acc2 male-acc-dat3 sell-fact-acc3 PIn−2-dat-dat4a artichoke-acc-acc4 speak-fact-caus2 elephant-acc-acc3 child-acc-dat3.

Weak bracket

Relative clauses can cause problems if their predicate differs from the underlying participle; this is notably the case with uncompounded modal verbs. The solution is similar to weak linking, only it is a bracket with an inner motivational or contextual case. The clause’s head has to be referenced with a pronoun in the dependent clause. (The English translation lacking the pronoun is a so-called reduced relative clause).

The underlying participle construction in these two examples is ‘the fed poodle’:
oRwxÌf avdíxty zèe.(works fine)the poodle [that] I have to feed
poodle-acc1 eat-fact-must-dat-acc2 PIn−3-nom-nom3a.
oRwxÌf lúlxty zeè ávy ciè zÌi.(the predicate is ‘want’: weak bracket necessary)the poodle [that] I want you to feed
poodle-acc1 want-mot-acc2 PIn−3-nom-nom3a eat-fact-acc3 PIn−4-dat-nom4a PIn−3-acc-dat4.

Often it is better to rephrase by an inversion or simply by leaving out the additional object per Rule Six.

láxt veè ávy ziè oRwxÌfi wÌcgy.The poodle I want you to feed is black.
I want you to feed the black poodle.
want-fact1 PIn−2-nom-nom2a eat-fact-acc2 PIn−3-dat-nom3a poodle-acc-dat3 black-acc-acc4.
oRwxÌf avlíxty zìe.the poodle [I] want you to feed
poodle-acc1 eat-fact-want-dat-acc2 PIn−3-dat-nom3a.

Headless relative clauses

On the risk of my readers getting funny ideas, I will call them ‘headless relatives’ for short. These are pronominal relative clauses that aren’t attributive because they have lost their referent (head, or predicate in Lemizh terminology) because it isn’t important. Thus, instead of saying ‘He found the thing that I wanted’ (attributive relative clause), we can say ‘He found what I wanted’ (headless relative). To do this in Lemizh, we omit the bracket’s unnecessary predicate along with the two case endings forming the bracket (one inner and one outer), pushing up its object by one level. This method works whenever a bracket’s predicate is unnecessary, even if we cannot form a corresponding English headless relative.

xÙlsk i lýxty cèe.He found the thing that I wanted.
search-fin1 make-acc-dat2 want-acc-acc3 PIn−4-nom-nom4a.
→ xÙlsk lýxty zèe.→ He found what I wanted.
search-fin1 want-acc-acc2 PIn−3-nom-nom3a.
dmìlt cOÌce klìpy oRwxÌfy.I know the person who stole the poodle.
see-cons1 human-acc-nom2 steal-dat-acc3 poodle-acc-acc4.
→ dmìlt klìpe oRwxÌfy.→ I know the one who stole the poodle. (‘nearly headless relative’)
I know the thief of the poodle.
see-cons1 steal-dat-nom2 poodle-acc-acc3.

In the second example we cannot get rid of the placeholder ‘the one who’ in English, wherefore I informally term this a ‘nearly headless relative’. Actually, we have been using them from the beginning – recall ‘the one who tells something’. This sentence is not the same as ‘I know who stole the poodle’: you can say ‘I know who stole the poodle but I don’t know the thief (personally)’. ‘I know who stole the poodle’ is an indirect question, which we will treat later in this unit.

The pronouns/adverbs ‘whoever, ‘wheneveretc. (not to be confused with the homonymous conjunctions) also introduce headless relatives. To capture the indefinite sense we can use an indefinite pronoun, which often results in a partitive bracket again.

xÙlsk gwÌi lýxtyn cèe.He found whatever I wanted.
search-fin1 any-acc-dat2 want-acc-partacc3 PIn−4-nom-nom4a.
qàzg gwÌy ganáryn lÌbe.I’m thinking of any [place] where the flowers sing.I’m thinking of wherever the flowers sing.
think-fact1 any-acc-acc2 sing-loc-partacc3 flower-acc-nom4a.
gìljd gwáy zìe.Whatever you do is fine.
good-cons1 any-fact-acc2 PIn−3-dat-nom3a.

Relative clauses and topicalisation

Since headless relatives aren’t (necessarily) brackets, the only difference to infinitive, gerund and conjunctional clauses seems to be the inner case, which is not a factive or affirmative. But then, the latter can be topicalised. So what is the difference? Nothing much, actually.

dmàt lusÌy dìy fOpysryfè dwÌwy.(relative clause)I see [the image of] Lucy, who gets/got a bottle from Father Christmas.
see-fact1 Lucy-acc-acc2 give-dat-acc3 FatherChristmas-acc-nom4 bottle-acc-acc4.
dmàt díy fOpysryfè dwÌwy.(nearly headless relative)I see the one who gets/got a bottle from Father Christmas.
see-fact1 give-dat-acc2 FatherChristmas-acc-nom3a bottle-acc-acc3.
(or topicalised gerund)I see the one getting/having got a bottle from Father Christmas.
dmàt dály fOpysryfè dwÌwy.(conjunctional clause)I see that Father Christmas gives someone a bottle.
see-fact1 give-aff-acc2 FatherChristmas-acc-nom3a bottle-acc-acc3.

Questions, imperative and exclamations

The rules of sentence grammar, specifically Rule Seven, imply that all sentences are actually declarative sentences (statements); but all languages have sentences in other moods as well. ‘Did you feed the poodle?’ and ‘Feed the poodle!’ do not claim reality of the feeding. ‘When did you feed the poodle?’ claims reality of the feeding only pragmatically, like ‘I see white mice’ suggests the existence of the mice.

Consequently, we have to paraphrase non-declarative sentences by moving the verb down to level 2 and introducing a new main predicate. For direct questions, this means that they become indirect questions with the main predicate pà. ‘I ask …’.

Direct questionIndirect question
When did you feed the poodle?I ask when you fed the poodle.
Did you feed the poodle?I ask whether you fed the poodle.

But how do we translate indirect questions?

‘What about’-questions

The simplest type of question consists of ‘ask’ as the main predicate plus some accusative object. ‘ask’ can easily be compounded since its nominative object is clear from context – it is myself. (See the modal verbs with irrelevant nominative.) Not compounding, i.e. using the word ‘ask’ as a standalone word, is more like an indirect question in English, while compounding, i.e. reducing it to a modifier, is more like a direct question.

htrÌy zìe. htrypÌ vìe.I ask about your sibling. What about your sibling?
ask-fact1 sibling-acc-acc2 PIn−3-dat-nom3. sibling-acc-ask-acc1 PIn−2-dat-nom2.
ìvy mlÌvy. ivpìl mlÌvy.I ask about the eater of the sweets. What about the one who has eaten the sweets?
ask-fact1 eat-dat-acc2 sweet-acc-acc3. eat-dat-ask-cons1 sweet-acc-acc2.

The translations of the accusative objects with inner non-factives as ‘ask about’ reflect what we have said about the preposition ‘about’.

The answer is constructed by referring to the question’s predicate with a pronoun. The pronoun normally doesn’t convey any useful information, so we get rid of it per Rule Six.

dmatnìl {fyè} veì dmÌyR.(to uncompounded question) I haven’t seen the asked-for one for a long time.I haven’t seen him for a long time.
see-fact-not-cons1 {PIIn−1-acc-nom2} PIn−2-nom-dat2 3/4-acc-dur2.
dmatnìl {fiè} veì dmÌyR.(to compounded question) I haven’t seen the asked-for eater …
see-fact-not-cons1 {PIIn−1-dat-nom2} PIn−2-nom-dat2 3/4-acc-dur2.

We can ask for any object in a sentence by inserting askÌ. ‘the asked-for thing’ and then making ‘ask’ the main predicate by inversion. This also works for factive objects (and even for words of deeper levels).

àv mlyvÌ pÌa. àvy mlÌvy. avpà mlÌvy.I ask about the action of eating sweets.How did you eat the sweets?
eat-fact1 sweet-acc-acc2 ask-acc-fact2. ask-fact1 eat-fact-acc2 sweet-acc-acc3. eat-fact-ask-fact1 sweet-acc-acc2.
fìlt {y}.(to compounded question)Fast.
fast-cons1 {PIIn−1-fact-acc2}.


Wh-questions are introduced with interrogative pronouns or adverbs (which look like the relative ones in English). As with relative clauses, different pronouns and adverbs correspond to different inner cases of the object: ‘Who has eaten the sweets?’ has the dative, ‘Why did you eat the sweets?’ the persuasive, ‘How (by which means) did you eat the sweets?’ the instrumental (in contrast to the factive question ‘How (in which way) did you eat the sweets?’ above).

This type asks for the identity of an object: in ‘Who has eaten the sweets?’, the content of asking (the thing asked about) is the identity of the eater, as opposed to ‘What about the one who ate the sweets?’, where the person asked about is the eater himself. Luckily, the word ‘identity’ is rather short in Lemizh: Ìd., the inner accusative of the verb àd. ‘give someone/something an identity’. ‘the identity of the eater’ has the eater as dative object, as it is the one who is given an identity. The pronoun in the answer typically forms a bracket.

Ìdy ìvi mlÌvy. ydpà ìvi mlÌvy.I ask about the identity of the eater of the sweets.I ask who has eaten the sweets. Who has eaten the sweets?
ask-fact1 identity-acc-acc2 eat-dat-dat3 sweet-acc-acc4. identity-acc-ask-fact1 eat-dat-dat2 sweet-acc-acc3.
htrybÌ {y}.The one given the asked-for identity is the sister.My sister.
sibling-acc-female-acc1 {PIIn−1-dat-acc2}.

Often it is sufficient to ask for a person or thing (‘what about’) as opposed to the identity so that we can omit yd-.

The interrogative pronoun ‘what’ can be translated with a bracket, ‘which’ with a partitive one.

ydpà Ìvi mlÌvy(n).I ask about the identity of the eaten thing, a sweet / from the set of sweets.What/which sweet did you eat?
identity-acc-ask-fact1 eat-acc-dat2 sweet-acc-(part)acc3.
yvpÌ mlÌvy(n).I ask about the eaten thing …
eat-acc-ask-acc1 sweet-acc-(part)acc2.

Ìd. is also used in indirect questions with main predicates other than askà., in which case it becomes important.

Mark the difference between nearly headless relatives and indirect questions:
xOìlj dèy mlyvÌ zìi.I have heard about the giver of the sweet.I have heard about the one who has given you the sweet.
hear-cons1 give-nom-acc2 sweet-acc-acc3 PIn−3-dat-dat3.
xOìlj Ìdyi mlyvÌ cìi.I have heard about the identity of the giver of the sweet.I have heard who has given you the sweet.
hear-cons1 identity-acc-acc2 give-nom-dat3 sweet-acc-acc4 PIn−4-dat-dat4.
dmìlt klìpe oRwxÌfy.I know the poodle’s thief.I know the one who stole the poodle.
see-cons1 steal-dat-nom2 poodle-acc-acc3.
dmìlt Ìdy klìpi oRwxÌfy.I know about the identity of the poodle’s thief.I know who stole the poodle.
see-cons1 identity-acc-acc2 steal-dat-dat3 poodle-acc-acc4.

Alternative questions (‘or’-questions)

We have already treated the inclusive and exclusive ‘or’, so alternative questions are not a problem.

jaxlixtpà nynn tfÌndin.What about the recipient of wanting to go, the valley and/or the forest?Do you want to go to the valley and/or the forest?
move-fact-want-dat-ask-fact1 valley-partacc-partdat2 forest-partacc-partdat2.
jaxlixtpà ryì nynwìn tfÌndin.What about the recipient of wanting to go, either the valley or the forest?Do you want to go to the valley or the forest [but we can’t do both]?
move-fact-want-dat-ask-fact1 one-acc-dat2 valley-partacc-partdat2 forest-partacc-partdat2.
nÌw {y}.[To] the valley.
valley-acc1 {PIIn−1-dat-acc2}.

Polar questions (‘yes/no’-questions)

Polar questions don’t ask for an object but for the predicate of the queried verb: ‘Are we going?’ — goày. ‘No, we aren’t going’. We cannot use an inversion to arrive at a construction parallel to the ones above, as the answer ‘No’ would then violate inversion ban. Instead, we use the general verb là. as a placeholder for the predicate.

Parallel to the answer ‘No’ we can form others such as ‘Well, I want to’.

lày jàxy nÌwi. lapà jàxy nÌwi.I ask about the predicate of going to the valley.I ask whether we are going to the valley. Are we going to the valley?
ask-fact1 do-fact-acc2 move-fact-acc3 valley-acc-dat4. do-fact-ask-fact1 move-fact-acc2 valley-acc-dat3.
nà {fÌy}.No.
not-fact1 {PIIn−1-acc-acc2}.
làxt {fÌy}.I want to.
want-fact1 {PIIn−1-acc-acc2}.

Note that compounding is compulsory for direct polar questions; otherwise the answer ‘yes’ becomes impossible. By the way, Ì. doesn’t just repeat the queried action as is the case in some other languages lacking a word for ‘yes’; it promotes its degree of reality from second to first level.

Topicalisation of goà. allows for the more specific queries ‘Are we going to the valley?’ and ‘Are we going to the valley?’.

Again, mark the difference in indirect questions:
wìlx jàxy nÌwi.I’ve been told about going to the valley.
speak-cons1 move-fact-acc2 valley-acc-dat3.
wìlx lày jàxy nÌwi.I’ve been told whether we are going to the valley.
speak-cons1 do-fact-acc2 move-fact-acc3 valley-acc-dat4.

Rhetorical questions

Rhetorical questions can be phrased like all other questions. Sometimes they are better translated as compounds with the verbs doubtà. ‘doubt’ or xràtx. ‘fear’.

kastxrátx vìe.I fear you are weeping.You are not weeping?
weep-fact-fear-fact1 PIn−2-dat-nom2a.


Requests and commands can be phrased directly with the verbs pràk. ‘request’ and dàxt. ‘must’, but also with various modal verbs and (rhetorical) questions, optionally compounded with weighting numerals to express different degrees of politeness.

avprák viè oRwxÌfi.I request that you feed the poodle.(Please) feed the poodle!
eat-fact-request-fact1 PIn−2-dat-nom2a poodle-acc-dat2.
avdáxt viè oRwxÌfi:You must feed the poodle!
eat-fact-must-fact1 PIn−2-dat-nom2a poodle-acc-dat2!
avlaxtcrá viè oRwxÌfi.I want you a little to feed the poodle.
eat-fact-want-fact-1/4-fact1 PIn−2-dat-nom2a poodle-acc-dat2.
avpá viè oRwxÌfi.What about you feeding the poodle?
eat-fact-ask-fact1 PIn−2-dat-nom2a poodle-acc-dat2.

The modal verbs làxt. ‘want’ and Ràks. ‘should’ can also express advice such as directions for finding the way: ‘Turn left!’ translates as ‘[You] want to turn left’ or ‘[I] suggest you turn left’.


Verbs such as fràx. ‘astonish’ translate exclamations.

fràx prìljOl: priljfràx:The beauty astonishes me!How beautiful!
astonish-fact1 beautiful-cons-psu2! beautiful-cons-astonish-fact1!

Tag questions

English tag questions (‘isn’t it?’, ‘do you?’) have a wide variety of uses. They can be actual questions (‘This is your poodle, isn’t it?’), rhetorical questions expressing a request or command (‘You’d better stop now, hadn’t you?’, ‘Do listen, will you?’); they can express politeness, emphasis, or irony; confidence or its lack, etc. A thorough discussion would probably require its own unit, so please use your imagination, possibly in combination with modal verbs, verbs of certainty, and weighting numerals.


We went to the valley where the flowers sing.
(with partitive, two possibilities)
the child that got a bottle vs. the child that took a bottleSolve
The child, which the tortoise assumes has got the bottle, is beautiful.
(Translate with and without a weak bracket.)
He found whichever poodle I wanted.Solve
Why did you eat the sweets? — Because they taste nice.
(Include the pronoun in the answer.)
Will you eat the sweets? — If I may.Solve
xèsk oRwxÌfy. (three possibilities)Solve
Why is it always possible to remove a bracket’s predicate from a sentence without breaking any dependencies?Solve