Roughly speaking, each Decalogue has two strands of cantillation, the טעם תחתון and the טעם עליון. Why is this only roughly true? Because in detail there are four strands: the printed tradition (p-trad) and the manuscript tradition (m-trad) differ in cantillation, yielding two different עליון strands and two different תחתון strands. (The p-trad is fading in popularity, but still visible in editions like Koren's Tanakh and Simanim's Tiqqun.) The most striking difference between the four strands is how they divide up the span אנכי…מצותי into chanted verses. (This span is typically identified as comprising the first two Commandments.)
I have run Richard Goerwitz's accent grammar checker against the m-trad strands (as represented by WLC); this page shows what happens when the checker is run against the p-trad strands.
There is no dual cantillation to detangle here. Whereas the Tiberian manuscripts (and WLC) pack both of their strands onto one set of letters, the Wikisource page עשרת הדברות בסיס/טעמים spells out all the strands separately — p-trad and m-trad alike — each as single cantillation. Every chanted verse (delimited by its own sof pasuq) is grammar-checked separately. The m-trad is the baseline — MAM's authoritative text, expected to be grammatical throughout — and the p-trad is the object of study.
| m-trad verses | p-trad verses | |
|---|---|---|
| Ex. תחתון | ✓ all 12 clean | ✓ all 13 clean |
| Ex. עליון | ✓ all 10 clean | ✗ 1 of 9 ungrammatical (verse 1) |
| Deut. תחתון | ✓ all 12 clean | ✓ all 13 clean |
| Deut. עליון | ✓ all 10 clean | ✗ 1 of 9 ungrammatical (verse 1) |
The תחתון is grammatical everywhere — both books, both traditions. In terms of verse boundaries, the p-trad תחתון differs from the m-trad in only one place: the p-trad תחתון ends its first verse at מבית עבדים (so it has one more verse than the m-trad, whose first verse runs on to על־פני instead). Both are grammatical. It also differs in some further details — several of them changes of cantillation that give the two strands genuinely different accent-grammar trees, but none that costs either its grammaticality (catalogued in an appendix below).
The עליון is grammatical in the m-trad but not in the p-trad: the p-trad עליון of each Decalogue has exactly one ungrammatical chanted verse — its opening one. That verse is not only an idealized strand's: real editions have it, as the transcription totals below say.
Every p-trad strand accents the Decalogue's אנכי…מצותי span differently from every m-trad strand. Two chanted words in the span do all the deciding — its signal words, עבדים and על־פני. Where a signal word has a silluq + sof pasuq, that strand ends a chanted verse there; where it instead has an etnaḥta or a revia, it runs on. The table below lays all four strands over four shared chanted words — אנכי (the span's shared start), then עבדים, על־פני, and מצותי (the two signal words and the span's shared end). Each row shows that strand's pointing at all four; a cell is red where its chanted word ends one of that strand's chanted verses. So every row starts green at אנכי and ends red at מצותי, and the reds between them are the signal accents at work — the strands' differing chanted-verse boundaries, read straight down the columns:
| mT | אנכי֙ | עבד֑ים | על־פנֽי׃ | מצותֽי׃ |
|---|---|---|---|---|
| mE | אנכ֖י | עבדֽים׃ | על־פנ֗י | מצותֽי׃ |
| pT | אנכ֖י | עבדֽים׃ | על־פנֽי׃ | מצותֽי׃ |
| pE | אנכי֙ | עבד֗ים | על־פנ֗י | מצותֽי׃ |
These are facts about the four strands as Wikisource defines them — idealized strands, each pure by construction. An actual edition need not follow one strand purely, and where it strays, the two signal words above cannot catch it. Simanim's Tiqqun, for instance, follows the p-trad תחתון's chanted-verse boundaries throughout, yet has the m-trad accents at the Shabbat commandment of its Deuteronomy appendix Decalogue. Catching that departure requires giving the Shabbat commandment its own signal words — three disjunctively accented chanted words, catalogued in the appendix below, where the two traditions differ in accents while agreeing on every chanted-verse boundary. The two sets of signal words have complementary jobs: עבדים and על־פני place a text among the four strands within אנכי…מצותי, while the Shabbat trio tells p-trad accents from m-trad accents at a commandment where a real edition can stray from its nominal tradition.
That departure by the Simanim Tiqqun is checked, not supposed. Twelve printed Decalogues — four each from Simanim's Tiqqun, Simanim's Tanakh and Koren's Classic Tanakh — have been checked accent by accent against the Wikisource strand each follows; the two satellite pages linked below report the verdicts, which are per Decalogue rather than per edition. Two results belong here instead, being facts about the strands rather than about one edition. All twelve follow their Wikisource strand's chanted verse boundaries without exception, whatever else they do. And at ובנך־ובתך, Simanim's Tiqqun and Koren each have two separately accented chanted words where the Wikisource p-trad עליון has one — two editions exchanging the same maqaf for a conjunctive, and against the Wikisource strand they otherwise follow.
One thing to settle first, since it decides what these verdicts mean. A maqaf belongs on the same scale as the accents, at the bottom of it: it separates the atom it sits on — one written word — from the next even less than a conjunctive accent does, binding the two into a single chanted word, the unit an accent marks. So where an edition has a maqaf on an atom and its Wikisource strand a conjunctive accent, or the other way about, that is one difference in how that atom is marked — an exchange one rung deep, counted once and not twice as a regrouping plus an accent. It is, though, the mildest difference there is, and that is what these verdicts measure: how far down the scale each Decalogue's agreement with that strand reaches — the chanted verse boundaries, then the disjunctive skeleton, then the conjunctives, then the maqafs.
Of these four strands, only the p-trad עליון is ungrammatical — the verdict above already said so — and the section below dissects its merged opening verse.
The twelve transcribed Decalogues have been through this same checker, chanted verse by chanted verse, and what that shows first is that following a strand and parsing are separate things. 10 of the 12 give their Wikisource strand's verdicts at every chanted verse — which for the 4 of them that follow a p-trad עליון means printing that ungrammatical opening verse. Simanim's Tiqqun does, in both of its main Decalogues, and Koren in both of its appendix ones: the merged first two commandments are in books in print, not only in an idealized strand. The remaining 2 are both Simanim's, and each is ungrammatical at a chanted verse where its Wikisource strand is clean, over a single mark. The Simanim Tiqqun's Exodus appendix Decalogue has a conjunctive its Wikisource strand does not; the Simanim Tanakh's Deuteronomy main Decalogue has a qadma where its Wikisource strand has a pashta. The Simanim page has both.
As the four-strands table above laid out, in the m-trad עליון the first commandment אנכי…עבדים is its own chanted verse, and לא יהיה לך אלהים אחרים begins the next — two separate verses, both of which are grammatical. The p-trad instead merges the first two commandments into a single verse. That one merged verse is what the grammar rejects. Shown stripped to letters and accents and divided at the p-trad תחתון's verse boundaries — one p-trad עליון verse (pE) spanning the five ordinary p-trad תחתון verses (pT) it merges:
| pE | אנכי֙…עבד֗ים | ל֣א יהיה־לך֩…על־פנ֗י | ל֣א תעשה־לך֣…לא֗רץ | לא־תשתחו֣ה…לשנא֑י | וע֤שה…מצותֽי׃ |
|---|---|---|---|---|---|
| pT | אנכ֖י…עבדֽים׃ | לא־יהי֥ה לך֛…על־פנֽי׃ | לא־תעש֨ה לך֥…לאֽרץ׃ | לא־תשתחו֥ה…לשנאֽי׃ | וע֥שה…מצותֽי׃ |
| pe | |||||
| pt | |||||
The single etnaḥta falls at לשנאי, leaving a first half crowded
with a segolta plus three separate
revia domains (at עבדים, על־פני, and
לארץ); the prose grammar cannot build it and returns a pashta_phrase error —
the same failure, at the same structural point, in both Decalogues.
The cause is the merged structure, not sheer length: Deuteronomy's p-trad עליון Sabbath verse runs 55 chanted words and parses clean, while this merged verse (51) does not. Keeping the two commandments as the m-trad's two separate verses (9 and 42) is exactly what lets them parse.
It is tempting to call the p-trad's arrangement the neater one. Its five תחתון verses nest perfectly inside a single עליון verse — a clean five-in-one — whereas the m-trad divides the very same chanted words into two עליון verses and four תחתון verses that do not nest cleanly at all: the m-trad תחתון's opening verse אנכי…על־פני straddles the עליון's break after עבדים, so neither strand's verses sit wholly inside the other's. In the gradient bars below, mE's first break lands inside mT's first verse, which runs on to על־פני — that is the straddle:
| mE | אנכ֖י…עבדֽים׃ | ל֣א יהיה־לך֩…על־פנ֗י | ל֣א תעשה־לך֣…לא֗רץ | לא־תשתחו֣ה…לשנא֑י | וע֤שה…מצותֽי׃ |
|---|---|---|---|---|---|
| mT | אנכי֙…עבד֑ים | לא־יהי֥ה לך֛…על־פנֽי׃ | לא־תעש֨ה לך֥…לאֽרץ׃ | לא־תשתחו֥ה…לשנאֽי׃ | וע֥שה…מצותֽי׃ |
| me | |||||
| mt | |||||
But the p-trad's tidiness is bought with an ungrammatical cantillation — the very one dissected above. Once you see the accents that buy the five-in-one, its neatness evaporates. This is not to say the tidier nesting could not have been cantillated grammatically; perhaps some other accentuation would have carried it off. That is a theoretical question I am not prepared to settle here.
All eight strands (two books × תחתון/עליון × m-trad/p-trad) are taken from the Wikisource base page עשרת הדברות בסיס/טעמים (revision 3025606, 2026-07-20, New York time), which every p-trad-vs-m-trad comparison table there transcludes. A handful of wiki templates are resolved to plain pointed text for accent-grammar parsing (legarmeh and paseq both conflated down to a Unicode PASEQ, ketiv dropped from ketiv-qere, setumah and petuḥah divisions dropped); the chanted verses are split at sof pasuq.
Source attribution: Hebrew Wikisource, under CC-BY-SA 4.0.
See also this document on the Decalogues in the Simanim Tiqqun — its main Decalogues follow the p-trad עליון and its appendix Decalogues the p-trad תחתון. It follows the p-trad's chanted verse boundaries without exception, and its Deuteronomy main Decalogue follows the p-trad in every accent; what it does not follow is every cantillation detail elsewhere — most consequentially at the Shabbat commandment, where its Deuteronomy appendix Decalogue has the m-trad accents. That page also covers the separately published Simanim Tanakh, which is m-trad in all four of its Decalogues: the same publishing house, the opposite tradition.
See also this document on the Decalogues in the Koren Tanakh — it has the p-trad תחתון in its main Decalogues and the p-trad עליון in its appendix ones, in Exodus and Deuteronomy alike. Two of its four Decalogues match their Wikisource strand mark for mark; the other two differ from theirs at the last rung only, where a maqaf on one side faces a conjunctive on the other.
See also this record of the printed editions cited at ובנך — the one word above at which two editions and the Wikisource p-trad עליון disagree. Whether Wikisource or the editions have it right is asked at MAM-basics issue #208, and that page holds the editions the issue's reply cites, with a crop of each page where one has been taken.
Both תחתון strands are grammatical, so the differences below do not bear on the grammaticality question this page asks, though several of these differences do change the accent-grammar tree. I gather these differences here only to make concrete the verdict's point that the strands differ in ways that never cost either strand grammaticality. The one structural difference — the p-trad ending its first verse at מבית עבדים, and the boundary accents on אנכי, אלהיך, מבית, and עבדים that follow from it — is the verdict's point and is not repeated here.
In Exodus, those four chanted words aside, the two strands part at exactly one more, the vocalization of תרצח:
| m-trad | תִּרְצָ֖ח | qamats |
| p-trad | תִּרְצַ֖ח | pataḥ |
Both have the same tipeḥa; only the vowel differs.
Those two vowels are the two halves of a pausal alternation: a great many Hebrew words take one vowel where the reading pauses and another where it runs on. The general rule (Yeivin §199) is that the pausal form goes with an etnaḥta or a silluq, and the contextual form with every other accent.
Two other chanted words of the Decalogue alternate the same way, and at both of them what decides is תחתון against עליון, not p-trad against m-trad: all eight strands have the pausal qamats in the תחתון and the contextual pataḥ in the עליון. There the rule accounts for the vowel, because the two strands accent those two differently. The תחתון ends a chanted verse at על־פני and puts its verse's main break at מתחת; the עליון runs on through both. In each pair below the תחתון comes first:
| תחתון | עַל־פָּנָֽי׃ | qamats | silluq |
| עליון | עַל־פָּנַ֗י | pataḥ | revia |
| תחתון | מִתָּ֑חַת | qamats | etnaḥta |
| עליון | מִתַּ֜חַת | pataḥ | geresh |
With תרצח, we have that same alternation, this time splitting the two traditions. In the עליון it is the second and last chanted word of the chanted verse לא תרצח, so it has a silluq and both traditions have the pausal qamats on it. In the תחתון it sits mid-verse under a tipeḥa, which by the rule above should take the contextual pataḥ — and that is what the p-trad has, while the m-trad keeps the pausal qamats. A pausal vowel under a disjunctive weaker than either etnaḥta or silluq is common enough for Yeivin to catalogue it; what the two traditions differ over here is only whether to do it at this chanted word.
So that is how the p-trad and m-trad of the תחתון differ in Exodus. In Deuteronomy, they differ more than in Exodus. Over the whole Decalogue the two strands part at 12 chanted words: the four boundary words already discussed above, one that differs in vocalization alone (the same תרצח, noted again below), and 7 that all fall within a single stretch of the Sabbath commandment. Those 7 are differences of cantillation, so unlike the Exodus vowel-swap they give the two strands genuinely different accent-grammar trees. Below is that stretch — the run of chanted words the two strands accent differently, everything before and after it being identical — shown in each strand, m-trad above p-trad, stripped to letters and accents, each line ending on a disjunctive accent. The 3 chanted words those lines end on are the Shabbat commandment's signal words — a shorthand for the stretch, not the whole of it:
| m-trad | ל֣א תעש֣ה כל־מלאכ֡ה |
| p-trad | לא־תעש֨ה כל־מלאכ֜ה |
| m-trad | את֣ה ובנך־ובת֣ך ועבדך־ו֠אמת֠ך |
| p-trad | את֣ה׀ ובנך־ובת֣ך ועבדך־ואמת֗ך |
| m-trad | ושורך֨ וחמרך֜ וכל־בהמת֗ך |
| p-trad | ושורך֤ וחמרך֙ וכל־בהמת֔ך |
The one difference outside that stretch is purely vocalization — vowel-pointing, not cantillation: the same תִּרְצָ֖ח / תִּרְצַ֖ח split as in Exodus.
Possibly the only p-trad Bible on the web is CTR: a web version of The Complete Tanach with Rashi, served from Chabad.org. Its Deuteronomy 5 has not only the p-trad תחתון's signal pair (chanted verses ending at both עבדים and על־פני), but also the p-trad תחתון accents in the Sabbath commandment. Its whole Decalogue, in fact, is the p-trad תחתון as Wikisource defines it, down to all but one conjunctive accent.
Its Exodus 20 is stranger. The accents are the עליון's throughout, silluq included — but it has a sof pasuq at every boundary of both strands: the עליון's nine and the תחתון's thirteen, sixteen marks in all. Seven of the sixteen are the תחתון's alone, and at those the עליון has no silluq, and neither does CTR: the seven chanted words before those marks have four revia, an etnaḥta, a segolta, and a zaqef. What those seven close could be called a sof pasuq span, but it could not be called a chanted verse. CTR's Exodus is the עליון's nine chanted verses with the תחתון's sof pasuq marks added — sixteen sof pasuq spans.
For more on CTR: see my review, which calls it “certainly the weirdest Hebrew Bible on the web, and possibly the worst.”