The German word nicht is akin to the on/off switch on a television set - using it is not much different. If you are watching a TV show, it is best to refrain from employing it:

(1a)

Ich schaue fern.
I watch TV.

If, however, you wish to stop watching the TV show, simply press the switch once:

(1.b)

Ich schaue nicht fern.
I don’t watch TV.

If you change your mind, simply press the switch again and the screen will turn back on:

(1.c)

Ich schaue nicht nicht fern.
It's not that I don’t watch TV.

The German word nicht is a means of negation, and it can be used repeatedly. Germans use this small word every day and it is one of the most frequently used words in the language. If you want to negate something, it may seem that there is nothing simpler than utilizing nicht and you're well on your way.

However, nicht is more peculiar than you might think, and that is what I aim to illustrate here.

Let's begin with the observation that nicht happily occurs below, but not above, indefinite object DPs such as jemand (someone), etwas (something), or ein Mann (a man):1

(2.a)

? John hat nicht jemanden angerufen.
John has not someone up-called
√ nicht>jemand: John didn’t call anyone.

√ John hat jemanden nicht angerufen.
John has someone <not> up-called
√ jemand>nicht: There is someone, whom John didn’t call.

(2.b)

? John hat nicht etwas gegessen.
John has not something eaten
√ nicht>etwas: John didn’t eat anything.

√ John hat etwas nicht gegessen.
John has something not eaten
√ etwas>nicht: There is something, which John didn’t eat.

(2.c)

? John hat nicht einen Mann angerufen.
John has not a man up-called
√ nicht>ein Mann: John didn’t call any man.

√ John hat einen Mann nicht angerufen.
John has a man <not> up-called
√ ein Mann>nicht: There is a man, whom John didn’t call.

The first sentences in (2.a) to (2.c) are not ungrammatical, but they do sound unusual when heard without appropriate context. Native German speakers can often distinguish between the first and second sentences in the examples given above. The pattern can be generalized as follows:

? [CP … Subject … [IP … nicht … Object … [ … VP … ]]]
√ [CP … Subject … [IP … Object … nicht … [ … VP … ]]]

Why does it sound weird when sentential negation appears in front of an indefinite object DP but not after it? Explaining this pattern is no easy task. In the following sections, I will consider three different explanations in terms of word choice, anti-licensing, and focus.

Bad word choice

At first glance, there appears to be an easy explanation for this phenomenon. Consider, for instance, the phrase nicht jemand (not someone), which is pretty clumsy. A German speaker would prefer to use the word niemand (no-one) instead.

We could, therefore, assume that the reason for the clumsiness of the examples in (2) is bad word choice: We shouldn’t negate positive expressions such as jemand (someone) or etwas (something), if negative expressions such as niemand (no-one) or nichts (nothing) could be used instead whithout changing the meaning of the sentence:

(3.a) nicht (not) + jemand (someone) → niemand (no one)

√ John hat niemaden angerufen.
John has no-one up-called
√ nicht>jemand: John didn’t call anyone.

(3.b) nicht (not) + etwas (something) → nichts (nothing)

√ John hat nichts gegessen.
John has nothing eaten
√ nicht>etwas: John didn’t eat anything

(3.c) nicht (not) + ein (a/n) → kein (no)

√ John hat keinen Mann angerufen.
John has no man up-called
√ nicht>ein Mann: John didn’t call any man.

The differences in word choice appear to explain the (slightly) ill-formed nature of the examples in (2), as opposed to the acceptable nature of the examples in (3).

The scopal readings preferred in these instances suggest that negation has a wider scope than the indefinite DP. It may therefore be hypothesized that negative elements are a composition of a negation phoneme and an indefinite determiner at the base level, allowing the negative item to c-command (or dominate) the indefinite.

A similar phenomenon occurs with disjunctive DPs in object position:

(2.d)

? John hat nicht London oder Paris besucht.
John has not London or Paris visited
? nicht>oder: John didn’t visit neither London nor Paris.

√ John hat London oder Paris nicht besucht.
John has London or Paris not visited
√ oder>nicht: John didn’t visit either London or he didn’t visit Paris.

The oddity of the first sentence can be attributed to poor word choice. If one wishes to establish a reading in which the negative item has wider scope, it is preferable to use the negative composite weder ... noch (neither ... nor) rather than directly negating oder (or). This will yield better results:

(2.e)

√ John hat weder London noch Paris besucht.
John has neither London nor Paris visited
√ nicht>oder: John didn’t visit neither London nor Paris.

I am not entirely convinced by this argument. There are at least two cases in which it fails to hold.

The first has to do with specificity. Sentences with clausemate negation and indefinite object DPs are typically ambiguous, although the inverse scope readings (the indefinite dominates negation) are often dispreferred. The inverse scope of indefinite object DPs, however, varies with the level of descriptiveness of the indefinite.

The likelihood of being interpreted specifically or non-specifically parallels with the degree of descriptional poverty or richness. Naked etwas (something), for instance, never has inverse scope. Sentences with "heavier" indefinites, such as etwas Bestimmtes (something particular) or etwas, das ich mag (something that I like), in object position are ambiguous between a specific and a non-specific reading and can be interpreted in two different ways:

(4.a)

√ Niemand hat jemanden, mit dem ich zur Schule ging, angerufen.
No one has someone, with whom I to school went, up-called

√ niemand>jemand: No one called anyone.
√ jemand>niemand: There is someone, whom no one called.

(4.b)

√ Niemand hat etwas Bestimmtes gegessen.
No one has something particular eaten

√ niemand>etwas Bestimmtes: No one ate anything.
√ etwas Bestimmtes>niemand: There is something, which no one ate.

This observation can be easily extended to all the examples in (2.a) to (2.c). As the descriptiveness of the indefinite increases, a wide scope reading becomes more prominent:

(2.e)

? John hat nicht einen Mann, den er gestern gesehen hat, angerufen.
John has not a man, whom he yesterday seen has, up-called

? nicht>ein Mann: John didn’t call any man he saw yesterday.
√ ein Mann>nicht: There is a man, John saw yesterday, and John didn’t call him.

(2.f)

? John hat nicht etwas, das ich mag, gegessen.
John has not something, which I like, eaten

? nicht>etwas: John didn’t eat anything I like.
√ etwas>nicht: There is something I like, which John didn’t eat.

If you, however, use another negative element instead of combining nicht with an indefinite object, you loose one of the readings:

(3.d)

√ John hat keinen Mann, den er gestern gesehen hat, angerufen.
John has no man, whom he yesterday seen has, up-called

√ nicht>ein Mann: John didn’t call any man he saw yesterday.
* ein Mann>nicht: There is a man, John saw yesterday, and John didn’t call him.

(3.e)

√ John hat nichts, das ich mag, gegessen.
John has nothing, which I like, eaten

√ nicht>etwas: John didn’t eat anything I like.
* etwas>nicht: There is something I like, which John didn’t eat.

This demonstrates that combining "not+indefinite" results in a distinct meaning from using an indefinite negative element such as nichts (nothing), niemand (no one), or kein (no). The latter transforms the specific reading of an indefinite element (if it is prominent) into a non-specific one.

If the indefinite expression is short, its lack of detail increases the probability of a non-specific interpretation, where no notable difference exists. However, as the descriptive richness increases, specificity grows, and the distinction between the two examples' meanings becomes apparent.

It is evident that the slightly problematic readings of (2.e) and (2.f) cannot be attributed to inadequate word choice. The negative element that is part of niemand, nichts, or kein serves as a marker for non-specificity, whereas simple negation (nicht) does not.

Secondly, the phenomenon of odd nicht is not restricted to indefinite DPs. It also affects other types of object DPs, such as definite descriptions, proper names, and pronouns. These expressions sound unnatural when positioned below sentence negation.

(5.a) definite descriptions: der Mann

? John hat nicht den Mann angerufen.
John has not the man up-called

√ John hat den Mann nicht angerufen.
John has the man not up-called

(5.b) proper names: Maria

? John hat nicht Maria angerufen.
John has not Maria up-called

√ John hat Maria nicht angerufen.
John has Maria not up-called

(5.c) pronouns: er

* John hat nicht ihn angerufen.
John has not him up-called

√ John hat ihn nicht angerufen.
John has him not up-called

Since there are no other simple negative expressions that could substitute for "nicht Maria" or "nicht ihn," which would be an unusual characteristic of any language, the odd interpretations cannot be attributed to poor word selection. Thus, there must be another explanation.

Polarity and anti-licensing

An idea is that the unusual readings are not a peculiarity of nicht, but rather stem from certain types of noun phrases not being permitted within the scope of negation.

It is widely recognized that many languages have polarity items, which are sensitive to the linguistic environment in which they are used.

A polarity item is a lexical item that is intrinsically connected to either affirmation or negation. More specifically, a positive polarity item (PPI) is associated with affirmation, while a negative polarity item (NPI) is linked to negation.

The context in which a polarity item is used is known as its licensing context. While an affirmative statement generally serves as a licensing context for a PPI, negation usually licenses an NPI. This implies that negative polarity items (NPIs) can only appear in clauses that are negative (in a sense that will be explained below), whereas positive polarity items (PPIs) can only occur in clauses that are affirmative.

The most common examples of polarity items are the English determiners any and some. As an NPI, any is acceptable under negation, but it results in ungrammaticality if used in a plain affirmative clause:

(6)

√ I didn’t call anyone.
* I called anyone.

Conversely, some is grammatical in affirmative clauses, but unacceptable in negative ones. When some is used with negation, the sentence is slightly ill-formed, but not ungrammatical, and a wide scope reading is strongly preferred. Crucially, some cannot be interpreted within the semantic scope of negation; only a reading in which the indefinite takes wide scope over negation is available:

(7)

? I didn’t call somebody.

* not>somebody: I called nobody.
√ somebody >not: There is somebody who I didn’t call.

The contrast between any and some is commonly explained by the notion that negation licenses NPIs like any, while anti-licensing PPIs like some. This concept has been extensively discussed in the literature on polarity items.

The discussion of polarity items has focused on identifying the factors that create a negative context. In the late 1970s, William Ladusaw, building on the work of Gilles Fauconnier, found that the majority of English NPIs are licensed in downward-entailing environments, a hypothesis known as the Fauconnier-Ladusaw hypothesis.2 Consider the following entailment:

(8)

John isn’t is a farmer. ⇒ John isn’t a sheep farmer.

Although normally having the more specific property of being a sheep farmer entails having the more general property of being a farmer, in negative contexts the direction of implication is reversed.

Normally, entailment moves upwards, from a smaller set (such as sheep farmers) to a larger set (such as farmers). However, in the case of (8), the entailment involves moving downwards, which is why it is known as downward entailment. Downward entailment can be defined as:3

A function f of type <σ,τ> is downward entailing iff for all x, y of type σ such that x ⇒ y: f(y) ⇒ f(x).

In a downward entailing context, you can safely make inferences from sets to subsets. The examples mentioned above give rise to downward entailing contexts:

(9)

? John hat nicht jemanden angerufen. ⇒ ? John hat nicht eine Frau angerufen.
(John didn’t call anyone. ⇒ John didn’t call any woman.)

? John hat nicht etwas gegessen. ⇒ ? John hat nicht einen Apfel gegessen.
(John didn’t eat anything. ⇒ John didn’t eat any apple.)

? John hat nicht einen Mann angerufen. ⇒ ? John hat nicht einen großen Mann angerufen.
(John didn’t call any man. ⇒ John didn’t call any big man.)

? John hat nicht London oder Paris besucht. ⇒ ? John hat nicht London besucht.
(John didn’t visit London or Paris. ⇒ John didn’t visit London.)

In German, disjunctions and indefinites may act as polarity-sensitive items (PPIs), meaning their acceptability in certain contexts depends on the scope of negation.

If negation has wide scope over the indefinite or disjunction, these expressions may be considered anti-licensed, resulting in ungrammaticality. However, if negation has narrow scope with respect to an indefinite or disjunction, they appear outside of the anti-licensing context and become acceptable, which is exactly what we observe.

I'm skeptical that such an explanation can succeed. The absence of licensing for indefinites or disjunctions under negation is not consistently observed in other downward-entailing contexts. Take for instance the case of nie (never), which also creates downward-entailing contexts:

(10)

√ John hat nie jemanden geliebt. ⇒ John hat nie eine Frau geliebt.
(John never loved anyone. ⇒ John never loved any woman.)

Here are parallel examples to those in (2.a) to (2.d), and all of them are grammatically correct:

(11.a)

√ John hat nie jemanden geliebt.
John has never someone loved
√ nie>jemand: John never loved anyone.

(11.b)

√ John hat nie etwas verheimlicht.
John has never something kept-secret
√ nie>etwas: John never kept anything secret.

(11.c)

√ John hat nie einen Mann angerufen.
John has never a man up-called
√ nie>ein Mann: John never called any man.

(11.d)

√ John hat nie Paris oder London besucht.
John has never Paris or London visited
√ nie>oder: John never visited Paris and never visited London.

We observe similar results with ohne (without), which also creates a downward entailing context:

(12)

√ John kam ohne Blumen. ⇒ John kam ohne Rosen.
(John came without flowers. ⇒ John came without Roses.)

The data exhibits the same pattern as we observed with nie:

(13.a)

√ John schwieg, ohne jemanden anzusehen.
John was-silent without someone to-glance-at

√ ohne>jemand: John didn’t glance at anyone.

(13.b)

√ John kam ohne etwas Besonderes.
John came without something special

√ ohne>etwas Besonderes: John came without anything special.

(13.c)

√ John verspätete sich, ohne einen Ausrede zu haben.
John came-late himself without an excuse to have

√ ohne>eine Ausrede: John came late without having any excuse.

(13.d)

√ John kam ohne Freunde oder Verwandte.
John came without friends or relatives

√ ohne>oder: John came without friends and without relatives.

Another relevant category is Strawson-decreasing operators, such as oft (often) or nur (only). These operators also create anti-licensing contexts for positive polarity items, but their relationship to downward entailment is less clear:

(14)
Only Jones is a farmer. -/-> Only Jones is a sheep farmer.

The invalidity of the inference in (14) is due to a presupposition that interferes with downward entailment.

In the case of only, it is typically assumed that the exclusive meaning component is asserted, while the truth of the prejacent (the clause without only) is presupposed. The inference from left to right is valid if we assume that Jones is a sheep farmer, but otherwise, we have a presupposition failure.

If we add the presupposition of the prejacent as an additional premise, we obtain a valid type of inference. For example, the sentence "Only Jones is a sheep farmer" presupposes that "Jones is a sheep farmer," therefore:

(15)

Only Jones is a farmer. & Jones is a sheep farmer. ⇒ Only Jones is a sheep farmer.

This leads to a generalization of downward entailment, known as Strawson downward entailment, named after philosopher Peter Strawson who was the first to bring presuppositions into the forefront of semantical analysis:4

A function f of type <σ, τ> is Strawson downward entailing iff for all x, y of type σ such that x ⇒ y and f (x) is defined: f(y) ⇒ f(x).

Strawson downward entailing contexts are created by various presuppositional items such as oft (often), nur (only), the antecedent of conditionals, and factive verbs like bedauern (regret), among others.

The following examples demonstrate that German indefinites and conjunctions can occur in Strawson downward entailing contexts without any licensing issues:

(16.a)

√ John hat oft jemanden angerufen.
John has often someone up-called

√ oft>jemand: It happened often, that John called someone.

(16.b)

√ Nur John hat einen Freund angerufen.
Only John has a friend up-called

√ nur>ein Freund: Only John called any friend.

(16.c)

√ Nur John hat London oder Paris besucht.
Only John has London or Paris visited

√ Nur>oder: Only John has been to London or Paris.

(16.d)

√ Wenn wir jemanden anrufen, dann sind wir verloren.
If we someone up-call, then are we doomed

√ conditional>jemand: If we call somebody or other, we are doomed.

(16.e)

√ Wenn wir einen Arzt oder einen Apotheker anrufen, dann sind wir verloren.
If we a physician or a druggist up-call, then are we doomed

√ conditional>oder: If we call a physician or a druggist, we are doomed.

(16.f)

√ Ich bedauere, dass John eine kluge Frau angerufen hat.
I regret that John an intelligent woman up-called has

√ bedauern>jemand: I regret that John called an intelligent woman.

These examples demonstrate that indefinites and disjunctions are generally licensed in downward entailing contexts and in Strawson entailing contexts. Based on this data and the preceding arguments, it appears that we can confidently conclude that neither indefinites nor disjunctions function as PPIs in German.

Instead, the slightly unacceptable readings where nicht (not) appears above the object DP are likely due to some idiosyncrasy of nicht and do not stem from anti-licensing. Thus, it appears that we are dealing with a specific peculiarity of nicht rather than a systematic feature of indefinite DPs or negative operators more broadly.

More data

Before exploring alternative explanations for the oddity of nicht, I would like to present three additional sets of data.

Firstly, quantifier phrases, plural indefinites, and bare plurals appear to be more comfortable in the scope of negation. When it comes to post-object nicht, however, the situation is reversed:

(17.a) quantifier phrases: alle Männer

√ John hat nicht alle Männer angerufen.
John has not all men up-called

? John hat alle Männer nicht angerufen.
John has all men not up-called

(17.b) quantifier phrases: mehr als drei Frauen

√ John hat nicht mehr als drei Frauen angerufen.
John has not more than three women up-called

? John hat mehr als drei Frauen nicht angerufen.
John has more than three women not up-called

(17.c) plural indefinites: zwei Männer

√ John hat nicht zwei Männer angerufen.
John has not two man up-called

? John hat zwei Männer nicht angerufen.
John has two man not up-called

(17.d) bare plurals: Kirschen

√ John hat nicht Kirschen gegessen.
John has not cherries eaten

* John hat Kirschen nicht gegessen.
John has cherries not eaten

Secondly, the appearance of nicht before an indefinite object DP is acceptable if nicht is stressed and thereby focused.

The following dialogues illustrate this point, with nicht appearing before the object DP in a focused position. The oddity of pre-object negation disappears, while sentences with post-object negation are unacceptable. (The capitalized letters indicate the focused element):

(18.a)

A: Du hast irgendjemanden angerufen! Gib’ es doch endlich zu!
You called someone or other! Go on, admit it!

B: Das ist nicht wahr! √ Ich habe NICHT jemanden angerufen!
That is not true! I have not someone up-called!

√ nicht>jemand: I didn’t call anyone.
* jemand>nicht: There is someone, whom I didn’t call.

B: Das ist nicht wahr! * Ich habe jemanden nicht angerufen!
That is not true! I have someone not up-called!

* nicht>jemand: I didn’t call anyone.
√ jemand>nicht: There is someone, whom I didn’t call.

(18.b)

A: Du hast doch etwas versteckt! Zeig’ es endlich her!
You hide something or other! Go on, show it to me!

B: Das ist nicht wahr! √ Ich habe NICHT etwas versteckt!
That is not true! I have not something hidden!

√ nicht>etwas: I didn’t hide anything.
* etwas>nicht: There is something I didn’t hide

B: Das ist nicht wahr! * Ich habe etwas nicht versteckt!
That is not true! I have something not hidden!

* nicht>etwas: I didn’t hide anything.
√ etwas>nicht: There is something I didn’t hide.

These examples demonstrate that focus is a crucial factor. When pre-object negation is focused, the oddity typically disappears. However, there is one exception with regard to pronouns: they need to be focused themselves to avoid ungrammaticality:

(19.a)

√ John hat NICHT jemanden angerufen.
John has not someone up-called

(19.b)

√ John hat NICHT etwas gegessen.
John has not something eaten

(19.c)

√ John hat NICHT einen Mann angerufen.
John has not a man up-called

(19.d)

√ John hat NICHT den Mann angerufen.
John has not the man up-called

(19.e)

√ John hat NICHT Maria angerufen.
John has not Maria up-called

(19.f)

* John hat NICHT ihn angerufen.
John has not him up-called

√ John hat nicht IHN angerufen.
John has not him up-called

Finally, it is worth noting that the oddity of pre-object negation also disappears when negation is related to certain sentence adverbials and particles, such as sofort (immediately), auf der Stelle (straight away), or doch (adverbial but/after all):

(20.a)

√ John hat nicht sofort jemanden angerufen.
John has not immediately someone up-called

(20.b)

√ John hat ja doch nicht Paris oder London besucht.
John has prt after all not Paris or London visited

(20.c)

√ John hat nicht auf der Stelle etwas geantwortet.
John has not straight-away something answered.

While similar patterns are observed with definite descriptions and proper names, this phenomenon does not hold true for pronouns:

(21.a)

* John hat nicht sofort ihn angerufen.
John has not immediately him called-up.

Furthermore, neither focus nor sentence adverbials resolve the issue of post-object nicht sounding odd when the object is a quantifying phrase, plural indefinite, or bare plural. The oddity remains regardless of whether negation is focused or accompanied by sentence adverbials or particles:

(22.a)

? John hat alle Männer NICHT angerufen.
John has all men not up-called

(22.b)

? John hat mehr als drei Frauen NICHT angerufen.
John has more than three women not up-called

(22.c)

* John hat Kirschen NICHT gegessen.
John has cherries not eaten

(22.d)

? John hat alle Männer nicht sofort angerufen.
John has all men not immediately up-called

(22.e)

? John hat zwei Männer ja doch nicht angerufen.
John has two man after all not up-called

(22.f)

* John hat Kirschen nicht auf der Stelle gegessen.
John has cherries not straight-away eaten

To summarize our findings so far, German nicht displays a rather complicated preference pattern when it comes to the placement of an object DP (ODP). Let's use NO as shorthand for 'nicht precedes ODP', and ON for 'ODP precedes nicht':

1. If ODP is a pronoun, then ON is preferred, and NO is ungrammatical. This can only be avoided by putting the pronoun in focus.

2. If ODP is a singular indefinite, a nominal disjunction, a definite description, or a proper name, then ON is preferred, and NO is odd but not ungrammatical. This can be avoided by focusing on either nicht or ODP, or when nicht is immediately accompanied by certain sentence adverbials.

3. If ODP is a quantifier phrase, a plural indefinite, or a bare plural, then NO is preferred, and ON is odd but not ungrammatical. This cannot be avoided by focus or other means (as far as we know).

4. Scopus usually follows overt word order, with specific indefinites being an exception. Word orders which involve odd readings, however, are not necessary for marking scopus. Special negative determiners such as niemand, nichts, kein, etc. can be used instead.

5. These patterns are specific to nicht, as they are absent when considering other downward-entailing operators, such as nie (never), ohne (without), nur (only), and so on.

I think the most plausible explanation for these patterns involves an interacton between focus and negation. Let me explain this.

Focus and constituent negation

Before considering focus, let’s start with the observation that multiple negations can appear in a sentence at the same time without canceling each other out.

This feature has been called negative concord. The basic idea of negative concord is that when there are multiple negative elements in a sentence, they combine to reinforce the negative meaning, rather than cancelling each other out as they would in standard grammar.

For example, in English, the sentence "I don't have no money" is an example of negative concord, where the two negative elements "don't" and "no" work together to mean "I have no money," rather than cancelling each other out to mean "I have some money."

In German, negative concord is known as "doppelte Verneinung" (double negation) and is a feature of some dialects, particularly those spoken in southern regions. In these dialects, multiple negative elements, such as "nicht" (not) and "kein" (no), are used to create a single negative meaning.

Consider for instance the following example of negative concord in Bavarian (which is a southern dialect of German):

(23.a)

√ I hob ned koa Wiaßwiaschd ned g'essa.
I have not no white sausage not eaten.
(I haven’t eaten any white sausage.)

Now consider Standard German, where negative concord is rarely used:

(23.b)

? John hat keinen Mann nicht angerufen.
John has no man not called-up
? (John did not call anyone.)
? (John called everyone.)

This sentence is not ungrammtical, but difficult to interpret. It has two possible interpretations: “John did not call anyone,” and “John called everyone.” According to the first interpretation, the double negative is interpreted as an example of negative concord, where the two negative elements keinen and nicht are working together to create a negative meaning.

In Standard German, multiple negative elements are typically interpreted independently. For instance, sentence (23.b) would usually mean "Peter did not fail to call anyone." When multiple negatives are used, they cancel each other out, and the sentence ends up meaning the opposite of what the initial negation implies.

The standard explanation for the distinction between the dependent (negative concord) and independent readings of negative items in a sentence is as follows: if multiple instances of negative items within a clause are bound together, with the first negative item dominating the others, then negative concord is observed. However, if they are not bound together, they are interpreted independently and cancel each other out.

Interestingly, if negative elements are emphasized, the independent reading becomes more prominent, making the sentence as a whole more acceptable:

(23.c)

√ John hat KEINEN Mann NICHT angerufen. Die Liste war lang, aber er hat sie alle erreicht.
John has no man not called-up. The list was long but he has them all reached.
* (John did not call anyone.)
√ (John called everyone.)

This again demonstrates that odd nicht cannot be attributed to word choice issues. More importantly, this example illustrates that even if a base position exists within a sentence where the negative particle nicht typically occurs, it may also appear in two or more positions within that same sentence if it is stressed and thereby placed into focus.

It can therefore be assumed that nicht has a preferred (unmarked) position, as well as several more marked positions, which give rise to focus-induced interpretations.

It appears plausible to assume that nicht usually appears within the IP-projection immediately above the VP-projection. In German matrix sentences, the verb appears in second position (V2) with an SVO order. If there is no object, adjective, or prepositional phrase, nicht appears at the end of the sentence (and the VP shell is empty apart from some bound traces):

(24.a)

√ John schnarcht nicht.
John snores not
(John doesn’t snore.)

[CP John-1 … [IP … schnarcht-2 … nicht [VP … t2 … t1 …]]]

It is a common assumption that the subject and the verb move out of the VP, which comprises almost all semantically significant lexical items (such as the verb, nouns, adjectives, and prepositional phrases) to an CP or IP-position for checking case or tense.

In German, when a verb consists of more than one element (such as including a prefix), the finite verb remains in the second position while the other parts go to the end of the clause. German grammarians refer to this as "Satzklammer" (sentence bracket). If the finite verb has a separable prefix, the prefix goes to the end, but nicht appears before it:

(24.b)

√ John ruft nicht an.
John calles not up
(John doesn’t call.)

[CP John-1 … [IP … ruft-2 … nicht [VP … an-t2 … t1 …]]]

Another example of an additional element is a past participle, which is used to form the present perfect tense. In this case, the finite verb is the auxiliary haben, which stands in the second position, while the past participle appears at the end of the clause, presumably within VP:

(24.c)

√ John hat nicht angerufen.
John has not called up
(John didn’t call.)

[CP John-1 … [IP … hat-2 … nicht [VP … t2 … angerufen … t1 …]]]

When in its base position, nicht is primarily utilized as a sentential negation operator. This notion is supported by the fact that a negated, non-stressed sentence can form a grammatically correct answer to a yes-no question:

(25.a)

√ Schnarcht John? Nein, John schnarcht nicht.
Snores John? No, John snores not.
(Does John snore? No, he doesn’t snore.)

However, that is not the only use of nicht. It can also be employed to denote a contrast to a focused element.

Focus is the highlighting or emphasis given to a particular word or phrase in a sentence, frequently to communicate new or essential information. German uses various methods to indicate focus, including word order, stress, intonation, or the use of specific particles or constructions.

The premise is that if nicht is linked to a focused element, it is used as constituent negation and implies the presence of significant new information. Let's refer to this as contrastive negative focus.

Here are some examples. The first illustrates a situation in which nicht is connected to a focus on the verb phrase. Although it (presumably) remains in its base position, a different interpretation is conveyed:

(25.b)

√ John SCHNARCHT nicht. Er rasselt wie ein Bär.
John snores not. He rumbles like a bear.
(John doesn’t snore. He rumbles like a bear.)

(25.c)

√ NICHT JOHN schnarcht. Maria schnarcht.
Not John snores. Maria snores.
(John does not snore. Maria does.)

(25.d)

√ John hat nicht JEMANDEN angerufen, er hat die ganze Firma verrückt gemacht.
John has not anyone up-called, he has the whole company crazy droven.
(John didn't call ANYONE, he drove the whole company crazy.)

(25.e)

√ John hat nicht TOM (IHN), sondern Brigitte (SIE) angerufen.
John has not Tom (HIM) up-called, but Brigitte (SHE) up-called.
(John didn't call TOM, but Brigitte instead.)

Thus, we can conclude that nicht has at least two distinct functions. When it is in its base position and no constituent is emphasized, it indicates sentential negation. However, if it is linked to a focused element, it can appear at various other positions, serving as constituent negation and signaling previously unrevealed new information that contrasts with the emphasized element.

In a sense, the listener is tasked with determining which word to substitute for the highlighted element in order to arrive at a true sentence.

Focus particles such as ja doch, nur, sogar, etc., are another way of emphasizing specific elements, and when combined with nicht, they can draw attention to alternatives:

(26.a)

√ John hat JA DOCH NICHT Paris oder London besucht. Er war in Rom.
John has prt after all not Paris or London visited. He was in Rome.
(John didn't visit Paris or London after all. He was in Rome.)

√ nicht>oder: John visited neither Paris nor London.

The phenomenon of odd nicht can be explained as follows: if nicht is not in its base position and is not bound to a focused element, the absence of stress makes the sentence sound odd. Thus, it is not strictly a grammatical issue but rather a complex phenomenon at the intersection of information structure, grammar, and prosody.

This is my final hypothesis. I did not examine complement sentences here. I did not discuss the positions in a syntactic tree where nicht appears. And I didn’t bother with focus-sensitive particles in any detail. However, one thing that still perplexes me somewhat is the phenomenon observed in the examples in (17). To illustrate, let me repeat one of them here:

(17.a)

? John hat alle Männer nicht angerufen.
John has all men not up-called

At first glance, it appears that nicht occupies its base position above the VP, and there is no specific element being emphasized in the sentence. However, despite this, (17.a) sounds awkward.

Neither stress nor focus-sensitive particles seem to alleviate the issue:

(22.a)

? John hat alle Männer NICHT angerufen.
John has all men not up-called

(22.d)

? John hat alle Männer nicht sofort angerufen.
John has all men not immediately up-called

One possible explanation is that quantifying phrases, plural indefinites, and bare plurals may not move out of the VP for some unknown reason.

If this were the case, nicht would not be in its base position, resulting in ungrammatical sentences. This could also clarify why focus does not improve the awkwardness of these sentences. However, this is purely speculative, and currently lacks any supporting data or evidence from my side.

In spite of my attempts to elucidate the matter, certain elements remain shrouded in mystery.

Epilogue

After finishing this article, I found myself pondering whether there might be alternative explanations that I had not yet considered. Given the helpfulness of ChatGPT on prior occasions, I decided to seek its assistance once again. However, despite my hopes, ChatGPT was unable to provide any novel insights on the matter, offering ideas similar to those already explored in the article.

One such possibility was the potential for poor word choice as a contributing factor:

The reason why this construction sounds slightly odd is that it is not the most natural or common way to express negation of a countable noun phrase in German. The more common way is to use the indefinite article "kein" (no) instead of "nicht einen", as in "John hat keinen Freund angerufen.”

I attempted to explain to ChatGPT why this wouldn't work, and to its credit, ChatGPT at least offered an apology for not being more helpful in that regard:

You are correct that using "kein" or other similar determiners would not be appropriate to negate the sentence "John hat nicht Klaus angerufen" (John did not call Klaus) in German. I apologize for any confusion caused by my earlier responses.

Now, let's consider the NPI and licensing idea:

The sentence "John hat nicht einen Freund angerufen" is an example of a negative polarity item (NPI) construction in German. NPIs are words or phrases that can only occur in negative contexts, such as "not" or "never" in English. In German, the NPI "nicht einen" (not a) is used to express negation of a countable noun phrase, such as "Freund" (friend) in the sentence you provided.

Upon highlighting that the given statement was nonsensical, I received a prompt response:

You are correct that the phrase "ein Freund" (a friend) is not an NPI in German. I apologize for the confusion in my previous response.

Lastly, ChatGPT presented the idea that focus, emphasis, and intonation could potentially play a role in the matter at hand:

The acceptability and naturalness of the sentence "John hat nicht Klaus angerufen" can depend on several linguistic and contextual factors, including:

1. Syntax and word order: The position of the negation particle "nicht" can affect the meaning and interpretation of the sentence. For example, "John hat Klaus nicht angerufen" (John did not call Klaus) has a different focus and emphasis than "John hat nicht Klaus angerufen" (John did not call Klaus).

2. Prosody and intonation: The way a sentence is spoken can also affect its acceptability and naturalness. The use of stress, pitch, and rhythm can signal different meanings and pragmatic effects. For example, the sentence "John hat nicht Klaus angerufen" may sound more acceptable with a rising intonation, which can signal a question or a confirmation-seeking tone.

3. Semantic and pragmatic context: The meaning and interpretation of a sentence can depend on the larger context of the conversation or discourse. For example, if the speaker has been discussing John's phone calls to various people, the sentence "John hat nicht Klaus angerufen" may be more easily understood and accepted than if it is presented out of context.

4. Speaker variation: Different speakers may have different preferences for syntactic structures and word choices in German, based on their dialect, regional variation, or individual linguistic style. For example, some speakers may prefer to use the indefinite article "einen" (one) in the sentence, as in "John hat keinen/einen Freund angerufen" (John did not call a friend).

In summary, the acceptability and naturalness of the sentence "John hat nicht Klaus angerufen" can depend on a range of linguistic and contextual factors, and may vary among different speakers and situations.

Based on the information presented, I believe this to be an accurate summary and will conclude the discussion at this point.

Larson (1988, 1990)1 and subsequently Chomsky (1993, 1995)2 proposed that the VP projection should be regarded as a VP-shell of the following form:

In this discussion, I will assume this thesis and explore whether the German clause-structure aligns with this framework. I find that fitting it into this structure is not straightforward. While it is possible with some effort, it appears rather cumbersome.

V-to-I movement

At first glance, Larson’s and Chomsky’s hypothesis about a universal VP structure seems applicable to the German clause structure. One could assume the presence of a head final IP above the VPs, and that V moves first to light v and then to the I-head (V-to-I movement). This results in the standard word order of embedded clauses in German:

(A)

[IP [vp Subj [v’ t [VP IO [V’ t DO]]]] [I’ Verb + I]]

This structure, however, does not account for complex VPs that include modal auxiliaries. The order of the modal-auxiliary-main verb complex in German, at least in finite clauses, is a mirror image of English. Consider the following examples of embedded German clauses with auxiliaries and modals:

(1)

… dass er es gegessen haben könnte.
… that he it eaten have may.
… that he may have eaten it.

(2)

… dass er betrunken war.
… that he drunken was.
… that he was drunk.

(3)

… dass er betrunken gewesen ist.
… that he drunken been has.
… that he has been drunk.

(4)

… dass er kommen wird.
… that he come will.
… that he will come.

These patterns are noteworthy. Modals or inflections select a bare infinitival verbal form, either an auxiliary or a main verb. An auxiliary such as haben, sein, or werden selects a past participle (known as Partizip II). This selection process involves choosing a complement and thus determining a hierarchical order. At first glance, V-projections appear to be head final in German. Therefore, the initial proposal (A) cannot be accurate.

Head-final VPs

Looking solely at the structure of the German clause, we can see that it exhibits a head-final VP-structure as exemplified in the following sentence:

(5)

.. dass die Frau geküßt worden sein wird.
... that the woman kissed been have will.
... that the woman will have been kissed.

While this proposal seems plausible, it contradicts a general argument put forth by Richard Kayne:3

… the human language faculty is in fact rigidly inflexible when it comes to the relation between hierarchical structure and linear order. Heads must always precede their associated complement position. Adjunctions must always be to the left, never to the right. That is true of adjunctions to phrases and it is true of adjunctions to heads.

This principle is known as the Linear Correspondence Axiom (LCA). It asserts that grammatical hierarchies in natural language must follow a universal order, namely specifier-head-complement branching order.

LCA is based on X-bar theory and posits that any phrase whose surface order is not specifier-head-complement must have undergone syntactic movements that disrupt this underlying order.

If Kayne's theory is correct, then the second proposal is problematic as it relies on a head-final structure. If German does exhibit a head-final structure on the surface, which is ruled out by LCA, we need to account for this through movement.

The first proposal, which involves simple movement to an IP-head, cannot explain German clause structure. It seems, however, that there are at least two more possible explanations for the fundamental clause structure of German, both of which involve V-movement.

Movement to functional heads above VP

Instead of assuming that German has a head-final VP structure, we could postulate the existence of functional heads above VP that display a reverse order, similar to what we see in English. Perhaps all Vs in German must move to these heads due to strong features or some other reason.

If we assume this hypothesis, the resulting structure would look like (B) with FP standing for some functional projection:

(B)

[FPpass-part …[FPpast-part …[FPbare-inf …[IP …[VPbare-inf …[VPpast-part …[VPpass-part …]]]]]]

Although this structure would be consistent with the LCA, there is little evidence in favor of this analysis. There are also several issues with this proposal:

(i) It is difficult to provide empirical evidence for it because the finite verb or a corresponding modal almost always appears at the end of the clause. There is nothing behind it that could be analyzed as an intervening constituent uncovering a head-V-movement.

(C)

… [FP verbal head [IP … [… ? … [VP trace]]]] …

(ii) If all functional heads (including IP) are V-projections, as this hypothesis assumes, then all arguments of the main verb and all adjuncts would also have to move, since they always appear in front of the Vs and never after them. In addition, VP-related adverbs like adverbs of manner always appear before the Vs and never after them. This is not consistent with what we observe in German:

(6)

... dass die Frau nur zaudernd / nicht geküßt worden sein wird.
*... dass die Frau geküßt worden sein wird nur zaudernd / nicht.

(iii) Assuming that all V-heads move out of VP would require many movements of arguments and adjuncts, which is not well-motivated and lacks empirical evidence.

Functional Projections within VP

A less ambitious proposal is to suggest that, while the IP is head-final, there are functional projections within the VP that the verbal heads move to. The structure could look something like:

(D)

[IP [VP [FPpass-part … [FPpast-part … [VPbare-inf … [VPpast-part …[VPpass-part … ]]]]]] I]

In this proposal, only the selected past and passive participles move, and the bare infinitive could move to I (out of VP), if there is no modal in that position. This solves problem (ii), but problems (i) and (iii) persist.

Additionally, all arguments of the main verb and all adjuncts would still have to move because they always appear in front of the Vs, never after them. Assuming all these movements seems unwarranted and lacks empirical support. Moreover, the clause in (7) is still ungrammatical:

(7)

*... dass die Frau geküsst worden sein nur zaudernd / nicht wird.

Oddities

I won't delve into the details of each proposal, but I would like to highlight two peculiarities that arise in all four of them.

(i) The first oddity concerns infinitival embedded clauses. In German, the equivalent of the Infl-position in English appears in front of the infinitive (here: zu sein):

(8)

Sie versucht, geküßt worden zu sein.
She tries kissed been to have.
She tries to have been kissed.

At first glance, this seems to support proposal (D). However, I would argue that instead of having a verbal position after Inf, the complex expression 'zu sein' is lexically generated and not syntactically derived properly, so it can sit with the bare infinitival in the I-projection.

To support this, I rely on composed infinitivals, i.e., verbs that have a prefix like zu, ab, auf, and so on. The infinitival marker 'to' appears between the prefix and the verb. For example:

(9)

Sie versucht, es abgeschrieben zu haben.
She tries it copied to have.
She tries to have it copied.

(10)

Sie versucht, es abzuschreiben.
She tries it to copy.
She tries to copy it.

I claim that the two words 'zu haben' in example (9) are actually not two components but one, which has the same position as abzuschreiben in example (10).

(ii) The second oddity, which is more complex, arises with CP-complement taking verbs, as these complements (and only these) always follow their V-head! To illustrate this with an embedded clause, we need to look at double embedded clauses:

(11)

... dass die Frau geglaubt haben wird, dass jemand sie küsste.
... that the woman believed have will that somebody her kissed.
... that the woman will have believed that somebody kissed her.

Although the following is also possible, it is clearly derived by topicalization movement:

(12)

... dass die Frau, dass sie jemand küsste, geglaubt haben wird.

This feature (that a CP-complement remains after all VPs) would also have consequences for the fourth proposal, which I repeat here:

(D)

[IP [VP [FPpass-part …[FPpast-part … [VPbare-inf …[VPpast-part …[VPpass-part …]]]]]] I]

If we stick to this structure, we would have to assume that the finite main verb doesn’t move to the I-head, because in this case, it would appear after the V-complement.

This alone wouldn't be too bad if we assume that the first position isn’t quite a VPbare-inf but a position where a base-generated finite verb could sit, so that its features could be checked by the I-head without movement (if it had weak features).

However, this assumption cannot be correct either. We can see this by looking at modals, which are not VP but IP-heads:

(13)

... dass die Frau glauben soll, dass jemand sie küßte.
... that the woman believe should that somebody her kissed.
... that the woman should believe that somebody kissed her.

(14)

*... dass die Frau glauben, dass jemand sie küßte, soll.
... that the woman believe that somebody her kissed should.
... that the woman should believe that somebody kissed her.

After considering all the factors, it appears that the German VP-system may actually align with the unpleasant third proposal. This would entail postulating numerous movements, including V-head movement, Aux-head movement, movement of all VP-related adverbials, and movement of most complements, except for CP-complements.

Furthermore, all DP-complements but no CP-complements would need to move, and even PP-adjuncts, which are supposed to appear in VP-adjunctions, would have to move out of VP.

Empirical testing of this structure is difficult, but the main reason for assuming it is the fact that CP-complements always follow the verb. While this is a reasonable feature of German clause structure, it raises questions about PPs, which are supposed to check their case internal to the PP.

Overall, this structure appears to be ad hoc and aesthetically unpleasing. To be truthful, it's a mess.