Triconsonantal Root Systems

[hwhatting]

it helps very much.

ah...I think I understand now:

if there is, say, kb-m, you know its from k-t-m

You mean "if there is kt-m"?

(and not a new root of kt-m) because there's no clustering.

is that a good description?

In principle, yes; I don't exactly remember how 4-consonant-roots work in this respect, but they are very rare, anyway, so the problem doesn't really arise practically.

[Rodlox]

it helps very much.

ah...I think I understand now:

if there is, say, kb-m, you know its from k-t-m

You mean "if there is kt-m"?

yes.

sadly, I manage to confuse myself. sorry.

[Bryan]

PIE may have evolved from a parent language based on bisyllabic roots, similar to most of the native roots in languages like Finnish or S?mi.

So what's this "native roots in languages like Finnish"? Please elaborate.

[Zhen Lin]

Query: are there any particularly common prefixes used to turn biconsonantal roots into triconsonantal ones?

Also: how did the various types of nominal derivation arise? What about the kind of derivation that happens nowadays, like the "software" pattern - how long has such "creative" derivation processes been around?

[Gremlins]

PIE may have evolved from a parent language based on bisyllabic roots, similar to most of the native roots in languages like Finnish or S?mi.

So what's this "native roots in languages like Finnish"? Please elaborate.

Apparently a vast number of PU roots have CVCV structure. Knock of the last vowel and you have a PIE-ish-looking root.

[hwhatting]

Also: how did the various types of nominal derivation arise? What about the kind of derivation that happens nowadays, like the "software" pattern - how long has such "creative" derivation processes been around?

If you mean "in languages generally", all I can say is that, for all what we know, the various types of nominal derivation may be around since people started to speak. Different languages simply have different preferences about which to use.
If you mean "in English" or "in IE languages", here is a simplified answer:
- suffixation goes back to PIE (as a process - most existing individual English suffixes are much younger)
- ablaut / apophony goes back to PIE (although I'd say in English it's not productive for nominal derivation anymore)
- umlaut (again not productive anymore for nominal derivationin English) goes back to North & West Germanic / Old English (in West Germanic it was probably still a phonetical process, it became morphological only in OE - or maybe even Middle English? I don't remember now)
- prefixation goes back to PIE composition; it became a separate process only in the IE daughter languages
- composition (the "software" type) goes back to PIE; the Germanic languages tended to develop it further, while other IE languages (e.g. Latin, Slavic) tended to limit its application

[Zhen Lin]

... I'm sure you've noticed that this is the triconsonantal roots thread? My query, of course, is about Arabic, Hebrew and other Semitic languages.

[hwhatting]

... I'm sure you've noticed that this is the triconsonantal roots thread? My query, of course, is about Arabic, Hebrew and other Semitic languages.

Oh, I'm on this board for so long that I stopped assuming that thread titles have any relationship with the matters discussed...

[Mecislau]

Ack, sorry. I never check this forum.

Query: are there any particularly common prefixes used to turn biconsonantal roots into triconsonantal ones?

Also: how did the various types of nominal derivation arise? What about the kind of derivation that happens nowadays, like the "software" pattern - how long has such "creative" derivation processes been around?

The nominal paradigms really aren't all that different than those you're see in IE langs. You've got normal abstract-noun endings, participles, etc. Modern, more specialized patterns are largely reinterpretations of earlier, more normal patterns. Note, for example, how terms like machshev "computer" (from Ch-Sh-B "think") is so similar to machshiv, the hif'il present participle of the same verb. Same prefix.

In the case of that software thing, the pattern comes from one of Hebrew's many agentives.

As far as I can see, there are no consonant clusters in roots, at least in Arabic, i.e. there are no roots that would be *kt-b-m (there's no /v/ in Standard Arabic). If you have a word X-kt-Y, with X and Y being consonants, this is either from a root x-k-t plus suffix Y or from a root k-t-Y with prefix X. It would normally not be from a root x-kt-Y (there are 4-consonant roots, but theyr're very rare).

Well, in Hebrew at least, four-consonant roots are definitely not rare. Five- consonant roots are, though they do occasionally appear.

However, you never see clusters in roots with three or less consonants. Biconsonantal roots are always C-C, and triconsonantal always C-C. Quadraconsonantal roots always treat the middle two as one (ie, C-CC-C). 5-cons always treat the first two as one (CC-CC-C).

So comparing the pi'el 3sg past masculine and feminine forms with D-B-R "spreak", Tz-L-Tz-L "ring", and F-L-R-T-T "flirt":

Masc: diber
Fem: dibrah

Masc: tziltzal
Fem: tziltzelah

Masc: flirtet
Fem: flirtetah

You can see how multiple consonants start to behave as one (regularly) as the number of consonants goes up. The added /e/ in the feminine forms are to break up the awkward cluster that would otherwise form.

[dragonprince99]

Out of curiosity, are there any rough estimates on the number of productive patterns in Hebrew/other similar languages, strict verbal forms aside?

[Nuntarin]

F-L-R-T-T "flirt"

One assumes this was a loanword. But why did they double the T?

[TomHChappell]

..

[Mecislau]

F-L-R-T-T "flirt"

One assumes this was a loanword. But why did they double the T?

To keep the F and L together and the R and T.

As I've said before, roots with more than three consonants start treating pairs as though they were single consonants. In quadraconsonantal roots, it's always the two middle consonants that get paired:

Tz-L-Tz-L "ring" > letzaltzel (with medial pair ltz)
T-L-P-N "telephone" > letalpen (with medial pair lp)
*F-L-R-T "(undoubled) flirt" > lefalret (with medial pair lr)

But "lefalret" really doesn't sound that much like "flirt" (to whoever borrowed it, at least), because the consonants are grouped together totally differently. To group them correctly, a redundant consonant had to be added to the end, because in 5-cons roots, the next pair to form comes from the first two:

F-L-R-T-T "flirt" > leflartet (with initial pair fl and medial pair rt)

This way, the same consonant groupings found in "flirt" are maintained.

[Mecislau]

Out of curiosity, are there any rough estimates on the number of productive patterns in Hebrew/other similar languages, strict verbal forms aside?

That's a tough question to answer. What do you consider strict verbal forms? I mean, the present tense forms of any verb can also be treated as nouns ("one who Xes") or as adjectives ("something that is Xing"). Many verbs also have passive participles that have acquired their own meaning (חשוב chashuv, originally "thought [of]", now means "important"). And then many word forms are tied to certain binyanim. Both כתיבה ktivah "writing" and התכתבות hitkatvut "correspondence" are gerunds, one derived from the pa`al form of K-T-B "write", one from the hitpa`el. Do you want to count these as different (because they form different nouns) or as the same (since they are both just gerunds, a simple nominalization, but from different forms of a verb).

My point is you're not going to be able to find a single number, because there's just far too many ways you can count what constitutes a different pattern. However, I can tell you that the number is not that large. I wouldn't put it higher than 40, probably much less.

[Gond]

I've read here that the spellings are standardized for the way the verbs and everything are written, because of the representation of the vowels....
I'm trying to make a conlang a bit like hebrew and I'm trying to find a way of representing the different tenses and verbal nouns with the same root that in a clear way but without using something like nikkudim (is it nikkudim or nikkud?), though I going to create something like that too... How exactly is the Hebrew spelling standardized? How does a child learn to write and read exactly when he's learning how to write without vowel points?

[TomHChappell]

I wouldn't put it higher than 40, probably much less.

I had earlier heard a number around 45. Lately though I've heard estimates as low as 25 and as high as 85. (Sorry, no sources.)

[Mecislau]

I've read here that the spellings are standardized for the way the verbs and everything are written, because of the representation of the vowels....
I'm trying to make a conlang a bit like hebrew and I'm trying to find a way of representing the different tenses and verbal nouns with the same root that in a clear way but without using something like nikkudim (is it nikkudim or nikkud?), though I going to create something like that too... How exactly is the Hebrew spelling standardized? How does a child learn to write and read exactly when he's learning how to write without vowel points?

It's nikkud (or niqud, however you prefer) - a mass noun.

Modern Hebrew uses a few consonant letters to double as a means of implying certain vowels. Vav (ו) (representing /v/, historically /w/) doubles for /o/ and /u/ and yod (י) (representing /j/) doubles for /i/ and in a few cases /e/. The letter alef (א), normally /?/, sometimes represents /a/ in loan words, and hey (ה) (normally /h/) indicates a word-final /a/ or /e/. However, these letters aren't all used whenever they can be, as there are very specific rules in place. However, they're not exactly that simple...

Here's a simplified list of rules for vowels, for reference (source: A Reference Grammar of Modern Hebrew - Edna Amir Coffin and Shmuel Bolozky), edited to remove the technical Hebrew-studies terminology:

1) Any consonant letter that functions as a vowel in fully-vowelled texts remains (representing historical consonants that dropped or vocalized). For example, the word הוא hu' is always spelt that way (HW') because there actually once was a /w/ there. The Proto-Semitic word was something like *húwa, IIRC.

2) Any /u/ is represented by vav.

3) Any /o/ represented by cholam in fully-vowelled texts (historical /o:/) is spelt with a vav. However, if the /o/ is represented by qamatz qatan or chataf qamatz (historical /O/), then:

• No vav is used if the qamatz qatan or chataf qamatz is found in all forms of the word.
• A vav is inserted if the qamatz qatan or chataf qamatz alternate with cholam in some of the word's forms.

4) The vowel /i/ is represented by yod in an open syllable, or in words whose base is an open syllable. It is not inserted:

• In a closed syllable that is closed in all forms.
• In some words who base form does not have /i/, but which changes to /i/ when declined.
• In causative verb forms (hif`il) when an initial /n/ has assimilated.
• Before /jo/ or /ju/
• In the prefix /mi/-
• In frequent function words, just as "if" or "with".

5) Generally /e/ is represented with yod when it comes from tzere (historical /e:/) provided the following conditions are met:

• The tzere replaces a basic /i/ before a gutteral consonant.
• When the tzere is maintained in all forms of a word.

The vowel /e/ from segol (historical /E/) is only represented with yod in the pattern heCeC.

6) Final /e/ and /a/ are indicated with hey, unless they came from words that historically ending with a glottal stop, in which case alef is used.

7) The above rules do not apply to native names, which retain their historical spelling.


All in all, though, the above rules work to ensure that different roots in the same pattern are generally spelt the same way, even when the vowels differ than in regular words due to the presence of gutteral consonants, glides, or other situations that historically changed the outcome of the vowels. This can make identifying the pattern and part of speech a given word is in quite easily, but very difficult to predict irregular vowel changes. For example, for both fairly regular plurals (such as קול qol "voice" > קולות qolot "voices") and very irregular plurals (such as עפרון `iparon "pencil" > עפרונות `efronot "pencils"), the plural forms are derived with exactly the same spelling!

Sorry, it's kind of hard to go into much more detail than that without getting fairly deep into the grammar...


As for children, as I understand they learn to read with the above spelling system (known as ktiv male') but marked with niqud as well. That means many vowels are redundantly marked, but it also makes it easier to just drop the niqud, since you're then left with the standardized Hebrew spelling.

Remember that Hebrew niqud are very archaic. There are 12-odd symbols used to represent only five vowels in the modern language. However, the 12-vowel niqud system is still used because it makes many irregularities in Hebrew grammar much more transparent as regular changes acting upon the pre-Biblical-Hebrew vowel system.

[Mecislau]

I wouldn't put it higher than 40, probably much less.

I had earlier heard a number around 45. Lately though I've heard estimates as low as 25 and as high as 85. (Sorry, no sources.)

That's just like I said - you can get all sorts of different numbers depending on how you count. It's just not as straightforward as you might expect.

[Gond]

Thanks, I think I've got it..., one day I shall learn Hebrew, but not yet.... I'll learn Arabic possibly as well..., but I'd learn every language in the world if that was possible xD

[Alioth]

Something's been bothering me.

Say you have a derived word (gonvah "pirated software" is what comes to mind). Its meaning is shaped from both the root and the vowel pattern. Then how do you inflect it, if inflecting means changing the vowel pattern? How would you say, eg, "one piece of pirated software", "several pieces of pirated software", "a collection of pirated software", "my pirated software"?

Granted, that example is of a derived noun. I can't think of any examples of derived verbs (and I'm not sure if they exist?), but the problem would be more severe in this case, because the binyanim are so big and complicated and baroque.

What crucial point am I missing here?

[Mecislau]

Something's been bothering me.

Say you have a derived word (gonvah "pirated software" is what comes to mind). Its meaning is shaped from both the root and the vowel pattern. Then how do you inflect it, if inflecting means changing the vowel pattern? How would you say, eg, "one piece of pirated software", "several pieces of pirated software", "a collection of pirated software", "my pirated software"?

There are two possible approaches here.

First of all, a bit of history. Hebrew and Arabic (I don't know enough about other Semitic languages to comment) took two different approaches in regards to their nominal systems. The Arabic nominal system is much more complex (in terms of irregularity) than Proto-Semitic, because Arabic developed very complex broken plurals. This means that regular plurals were largely replaced by collective nouns, which were then reanalyzed as plurals. For example, kitaab "book" lost its original plural, which would have been a regular affixial formation like *kitaabiima, in favor of kutub, which was the collective form of "book". This system, probably starting with nouns commonly found in collectives (things like "sheep" > "flock"), soon spread by analogy throughout Arabic. Proto-Semitic had a huge collection of collectives, so in a perfect world, every noun vowel pattern in Arabic would have its own plural vowel pattern. This does happen to be true in many cases, but since Arabic is a natural language, you've all all sorts of exceptions.

Now, Hebrew. Hebrew did inherit collectives too, but they never became nearly as productive as in Arabic. Many latter fell out of use, although some are still around: sa`arah "a hair", se`arot "hairs", se`ar "hair (collective)". Instead, Hebrew generalized two plural endings, -im for masculine nouns and -ot for feminine (although this isn't 100% the case). However, after this Hebrew went through what is perhaps the most complex vowel shift in any Semitic language (where every vowel could basically become every other vowel in certain situations, literally), and this included a lot of vowel weakening and dropping. However, nouns with the same vowel pattern generally underwent the same changes (since Hebrew, like other Semitic languages, underwent very few consonant changes). The result is actually something similar to Arabic's irregularities, though not nearly as dramatic (but still a pain to learn). So in modern Hebrew, when you're given a noun, you can generally predict exactly what changes are going to occur when you add a plural ending. Let's look over the changes involved in a pair like Hebrew melech "king" and mlachim "kings".

Singular:
PS *malkum
malk (Loss of case ending)
málek (Epenthetic vowel to break up cluster)
melek (a > e vowel change)
melech (lenition of /k/ to [X])

Plural:
malkíma
malkím (Loss of case ending)
málekím (Analogical change mirroring singular)
m@lákím (Weakening of initial vowel to schwa due to its antepenultimate position, /a/ lengthens in compensation)
m@láchím (Lenition of /k/)
mlachim (Loss of schwa)

(I'm not sure I got all of those changes right, but you get the idea)

The result is any noun coming from PS *CaCCum will have a modern Hebrew singular reflex in CeCeC, and from *CaCCíma in CCaCim. Analogy then helps spread this to other nouns that may not have undergone the expected changes for whatever reason.


So, in answering your question about plurals, the answer is that basically every singular noun pattern will have its own plural pattern. In Arabic, these will be very different. In Hebrew, the plural stem (without the endings) tends to be quite similar if not identical to the singular, and whatever changes do occur are generally predictable). So the plurals are all tied to the vowels of the singular.

The same goes for collectives, but just be aware the collectives are not all that productive in modern Semitic languages.

Now, possession. In Arabic and Hebrew this is actually a lot simpler than you'd probably expect. But first, we need some history again.

Semitic languages, in addition to case, have a quality known as "state". Different Semitic languages have different states. Every Hebrew noun has three, known as the absolute, definite, and construct. Other languages can have more or less, but the construct you'll see in many of them (if not all). If you want me to go into detail about what states are, just ask, but I'm not going to do it now.

Now, the construct state is used to form possessive relationships. Hebrew examples are actually a lot simpler here, because Hebrew has completely lost cases in favor of states, while Arabic maintains both.

Basically, every possessed noun has its own construct singular and construct plural forms. Again, like the plurals, these depend heavily on the vowel pattern of the absolute singular (the citation form, like "melech" earlier). Let's take a noun like ben "son". Here are its absolute and construct forms:

AbsSg: ben
AbsPl: banim
ConsSg: ben
ConsPl: bney

Here the construct singular and absolute singular are identical, but they don't have to be. Now, the construct noun represents a possessed noun. The possessor will be a following noun in either the absolute or definite states: ben ha'ish "the man's son", bney ha'ish "the man's sons".

Now, personal possession (my, your, etc) is also logically formed from the construct. For masculine nouns, you can get the construct stem by removing the -ey from the plural. For feminine nouns, the construct stem is the same as the construct singular (see example below). So, to say "my son", you have a theoretical construction like *bn(ey)- + 'ani. Of course, you can't actually leave the full pronoun like that. The pronoun reduces to its enclitic form, -i, leaving you with bni "my son". And so that continues - to form all of the other possessive forms, you just add their endings to the construct stem: bno "his son", bnah "her son", etc (there are a few irregularities at times, but you don't need to worry about those). The plurals are formed with a different ending, which historically comes from that -ey plural construct ending plus the pronominal endings: bnay "my sons", bnav "his sons", bneha "her sons", etc.

Now, using your gonvah example. This is a feminine noun, so it forms the construct a little differently, but the principle's the same:

AbsSg: gonvah
AbsPl: gonvot
ConsSg: gonvat
ConsPl: gonvot

Now, as I said, for feminine nouns the construct stem is the same as the construct singular, so you can just add the endings, and voilà! gonvati "my pirated software", gonvato "his pirated software", and so on. (Just a little warning, though - this word is a recent creation and still rather slangy (think "warez"), so you're never going to see possessive suffixes used on it. In Hebrew, possessive suffixes tend to only be used in formal registers, or on a small set of common nouns such as family members, so you'll never see possessive suffixes on a slang term).

Whew, I really did not expect to write that much, but I sort of had to detail exactly how Hebrew does it, since that's what you're asking. Of course, this certainly isn't the only way to do it. You clearly don't need to have something analogous to the Semitic construct to be able to form triconsonantal possessives, but that's just how the Semitic languages do it.

If anythings still unclear, please ask.

Granted, that example is of a derived noun. I can't think of any examples of derived verbs (and I'm not sure if they exist?), but the problem would be more severe in this case, because the binyanim are so big and complicated and baroque.

Again, each binyan has its own individual paradigm. You're not going to confuse which verb is in which binyan because all of their forms are different. Here, I'll give you the root *K-T-B "write" in the present masculine singular of each binyan. You can see they're all different:

Pa`al (active): kotev "write"
Nif`al (passive): nichtav "be written"
Pi`el (intensive): mechatev "engrave, write a lot"
Pu`al (passive intensive): *mechutav (I don't think this form of KTB is actually attested, but I'm not sure)
Hif`il (causative): machtiv "dictate"
Huf`al (passive causative): muchtav "be dictated"
Hitpa`el (reciprocal): mitkatev "correspond"

So you've got the same root, but each binyan has its own pattern.

Now, roots with problematic consonants, such as Y or V (historical *W), can mess with the paradigms and cause them to look a whole lot more similar, but they still do not merge. Like the nouns, each binyan has a specific pattern for specific irregularities, so, for example, each binyan has its own (separate) pattern, known as a gizrah paradigm, for roots beginning with *Y. Just for an example, here's Y-D-` "know":

Pa`al: yodea` "know"
Nif`al: noda` "become known"
Pi`el: meyadea` "inform"
Pu`al: meyuda` "be informed"
Hif`il: modia` "make known"
Huf`al: muda` "be publicized"
Hitpa`el: mitvadea` "reveal oneself"

Now, I'm not entirely sure what Arabic does, but it's probably not too far off from this, and likely a lot less baroque. Arabic's known for making its nouns a lot more complex while simplifying its verbs, and Hebrew for making its verbs a lot more complex while simplifying its nouns.



Sorry for the long post!

[Alioth]

Thanks! That does clear things up a lot.

[Wiseblood]

Okay, first post here - I found this topic so interesting that I had to get an account to participate.

First off, Maknas, thank you for a truly interesting discussion/analysis of Hebrew morphology (I found the software and disease examples particularly interesting). I've always been aware of the fact that Arabic and Hebrew were very similar languages, I don't think I'd ever realized just how similar they were until I started reading your posts.

In any case, two little nitpicks about the origin of segolates:

*melex might not have been the best choice for this specific explanation, as the Arabic (and I believe the Akkadian) cognate, malikun already has a vowel between C2 and C3. I believe *kalbum (dog) may have been a better choice. Actually, the fact that both types of nouns seem to function the same way in Modern (and I'm assuming Biblical) Hebrew is a testament to the power of sound change and the force of analogy in semitic languages.

The (masculine) nominative plural ending for PS was probably *-uuma or *-uum. This is based on classical Arabic's ending, -uuna, and Akkadian's ending, -uu. The oblique ending was probably *-iima or *-iim. Again, this is based on Arabic -iina and Akkadian -ii. As you can see, the Hebrew plural almost certainly derives from the oblique form (if you already knew this, please feel free to call me an idiot). The final -a* may have never existed; so far as I know, CA is the only language it's attested in, and there it may have simply been inserted to conform to CA phonological constraints (a word can't end in an extra-long syllable). Even if it was there, it probably had very little to do with case.

Again, these are silly little nitpicks - feel free to call me a jerk for pointing them out/disputing your analysis.

What really struck me about these forms and indeed, many of the other changes in Hebrew you describe, is how similar they are to the changes that happened to many modern Arabic dialects.

An example from my own dialect:

CA: kalbun (dog)
kalb (loss of case ending - seem familiar?)
kalib (epenthetic vowel breaking up final consonant cluster - eerie, isn't it?)

This has the net effect of making malik and kalib look exactly alike in Arabic too.

Unfortunately, I can't give a plural example, as it's a broken plural. Actually, for the life of me, I can't think of *any* word of the types CaCC/CaCiC with a sound plural (not to say there aren't any). However, watch what happens to the active participle of 'KL (to eat).

Fem
CA: 'aakilatun
'aakilah (loss of case endings/final unstressed syllable/vowel)
'aaklah (unstressed syllables disappear)
'aklah (long vowels shorten before cononant clusters)
of course, as per Hebrew, that (t) reappears in the construct state.

Masc Plur
CA: 'aakiluuna
'aakiluun (loss of case endings/final unstressed syllable/vowel)
'aakiliin (plural form transitions to oblique with the loss of case)
'aakliin (unstressed syllables disappear)
'akliin (long vowels shorten before cononant clusters)

Okay, that ended up being far longer than I intended. Out of curiosity, will these fora explode if I post using unicode IPA/latin extended?

[Drydic]

Okay, that ended up being far longer than I intended. Out of curiosity, will these fora explode if I post using unicode IPA/latin extended?

They shouldn't. A lot of fonts don't have any of the underdotted consonants, so might just want to use combining diacratics for those. (great post btw)

[hwhatting]

An example from my own dialect:

Which one is that? (from your example, it doesn't look like Lebanese - the only dialect I know a bit about).