Semantics vs agreement: What determines the interpretation of German bracketing paradoxes?

AMLaP 2024, September 5th 2024

Anna Prysłopska

anna.pryslopska@ling.uni-stuttgart.de

University of Stuttgart
Institute of Linguistics

Titus von der Malsburg

titus.von-der-malsburg@ling.uni-stuttgart.de

University of Stuttgart
Institute of Linguistics

German nominal compounds

der Obstbaum the fruit.tree

das Obst the fruit

der Baum the tree

German nominal compounds are a famously productive and nearly unrestricted class of words. Despite their complexity, they are easy to understand because they’re composed of lexemes that typically correspond to existing words. Obstbaum decomposes in Opst and Baum.

Adjective attachment in German

[ [ keltisches [ Fürsten]grab ]]

Celtic tomb of a prince

Tomb of the Celtic prince

Not compositional?

Same syntactic bracketing, different semantic bracketing, seemingly against the principle of compositionality.

Adjectives are prenominal and must morphosyntactically match the last noun in the compound (unlike in English).

When nomimal compounds are modified by adjectives, the adjective takes on the moprhological properties of the compound’s head. The head nouns is the last noun in the compound (i will call it the second noun/N2 throughout the talk).

Compounds together with adjectives can be interpreted in one of two ways. In the typical case, the adjective modifies the head noun = celtic attaches to the tomb (nationality of the prince is not important)

In some well attested cases the adjective can attach to the first noun(s) in the compound = tomb of a celtic prince.

Both constructions have the synactic bracketing on the left, but pradoxically their semantic bracketing is different. Principle of compositionality?! Non compositional?

These constructions seem not compositional, they used to be denounced as ungrammatical, but they are still very common.

Some phrases ambiguous or preferentially interpreted as a bracketing paradox

steiniger Fahrradweg stony bike.road
→ canonical reading, control

königliche Hochzeitsfeier royal wedding.celebration
→ canonical and bracketing paradox readings, unbiased

saurer Kirschbaum sour cherry.tree
→ bracketing paradox reading, N₁ biased

?? dreiköpfiger Familienvater 3-headed family.father
→ the father of a family of 3 vs. Cerberus, N₁ biased

Some phrases are clearly N2 or N1, others are somewhere in between. There is also no regularity to how natural or acceptable the bracketing paradox phrases are

How do we model these phrases?

How are they processed?

There is barely any psycholinguistic research on this phenomenon!

We asked ourselves: how do we model these phrases and how are they processed?

bad language, word play,

pragmatics, plausibility,

world knowledge, semantic

transparency, context, language

economy, adjective type,

compound lexicalization,

semantic compatibility,

morphosyntactic agreement

between the adjective and nouns

Experiment 1

Do semantic and morphosyntactic compatibility between the adjective and the individual nouns affect the acceptability of the entire phrase?

In the first experiment, we investigated whether the acceptability of the whole adjective-compound phrase is affected by the the semantic and morphosyntactic compatibility between the adjective and the individual nouns

Method

We gathered stimuli from literature and newspaper articles

204 adjective + nominal compound pairs from literature and newspaper articles

Canonical, unbiased, and bracketing paradox compounds

Excluded constructions where the adjective didn’t match either noun → unlikely to be used.

saurer Kirsch sour cherry

AN1

saurer Baum sour tree

AN2

saurer Kirschbaum
sour cherry.tree

AN1N2

Schmidt (1993) made empirical studies of similar phenomena, using some of Bergmann’s examples

3 lists, 36 Prolific participants, 3 scales

Predictions: Semantic compatibility

Strictly compositional view: N₂ is the head of the compound.

As AN₂ ratings go ↑↓, so should AN₁N₂
High match → control/canonical composition

N₁ should not impact the AN₁N₂ ratings
High match → bracketing paradox potential

Results: Semantic compatibility

Strictly compositional view: N₂ is the head of the compound.

As AN₂ ratings went ↑↓, so did AN₁N₂
High match → control/canonical composition

N₁ did impact the AN₁N₂ ratings
When AN₁ and AN₂ ratings were high, compound rating was reduced.

Predictions: Morphosyntactic agreement

When the adjective matches the N1 and N2, this could make readers more likely to accept a BP interpretation. Aided by morphosyntax. We annotated the agreement between the adjective and N1.

schwere Unwetter das schwere Unwetter schweres Unwetter

The adjective’s morphosyntactic features must match N₂.

When the adjective morphosyntactically matches N₁ → bracketing paradox potential

schwerefem/neut/pl	Unwetterneut/pl	warnungfem	✔
severe	weather	warning
saurermasc	Kirschfem	baummasc	✘
sour	cherry	tree

Results: Morphosyntactic agreement

The adjective’s morphosyntactic features must match N₂.

The morphosyntactic match between the adjective and N₁ had no effect.

Discussion

Both nouns have a positive influence on the compound’s acceptability.

Their effects are not strictly additive.

When adjective 🗲 head noun, the first noun becomes an attractive target for the adjective.

semantic compatibility ✔ morphosyntactic agreement ✘

→ neidischer Zickenkrieg jealous bitch.war

Experiment 2

Which noun in a compound is modified by the adjective?

Next, we asked ourselves which noun does the adjective ACTUALLY attach to

answer options: n1, n2, unclear (may opt out of interpreting altogether)

235 items in one condition, 20 Prolific participants

saurer Kirschbaum sour cherry.tree

Predictions

Grammatically, the adjective is always attached to N₂.

N₂ is chosen more often than N₁

N₁ is chosen more often than N₂

N₁ is chosen as often as N₂

? is chosen often

Adjective matches N₁ morphosyntactically

canonical ₁

bracketing paradox ₁

ambiguous ₁

no interpretation ₁

adjective attaches more easily to N₁

The semantic compatibility ratings from Experiment 1 should align with the selected attachment site.

Results: Attachment preference

Our corpus consisted of canonical and BP examples, so not unexpected to find N2 attachments

N₂ was chosen overall more often than N₁

N₁ was sometimes chosen more often than N₂

N₁ and N₂ were similar

? is chosen often

grammatical, canonical ₁

ungrammatical, bracketing paradox ₁

ambiguous or bracketing paradox? ₁

no interpretation ₁

Predictions

Attachment and AN₂ rating

The compatibility ratings should match the attachment site.

Results

Attachment and AN₂ rating

The compatibility ratings match the attachment site.

Adjective matched N₂ (semantically) reduced N₁ attachments.

Predictions

Attachment and AN₁ rating

Adjective N₁ semantic match should increase N₁ attachments.

Results

Attachment and AN₁ rating

Adjective N₁ semantic match increased N₁ attachments, especially when adjective and N₂ match was poor.

Results

Morphosyntactic match

Morphosyntactic match between the adjective and 1^st noun had no effect.

Discussion

Participants did not opt out of interpreting the phrases.

Most phrases were judged in accordance with grammar, but N₁ was a viable attachment site for a large group of phrases.

The participants disagreed on the interpretation of a substantial group of phrases.

The results are in line with Experiment 1.

semantic compatibility ✔ morphosyntactic agreement ✘

Experiment 3

We felt the phrases sounded better when embedded in a sentence or context.

Im Elternabend berichtete
der dreiköpfige Familienvater von seinen Erfahrungen im Zwiespalt zwischen Familie und Beruf.

Task: Rate the phrase.

Same methods as in Experiment 1, only one AN₁N₂ condition.

At the parents' evening,
the father of three reported on his experiences of the conflict between family and career.

Predictions

When bracketing paradoxes are embedded in a sentence, their rating will improve.

Results

When bracketing paradoxes were embedded in a sentence, their rating slightly improved.

Conclusion

The bracketing paradox interpretation is influenced by semantic rather than morphosyntactic agreement.

Acceptability ratings and attachments seem to be determined largely by semantic factors.

Bracketing paradoxes persist despite defying grammatical and strictly compositional principles.

Strong semantic cues can override grammatical constraints.

Compositional processing can be suspended to fulfill communicative goals.

Thank you

Future work

Väter mit einem Kind und solche mit mehreren Kindern erhalten unterschiedliche staatliche Unterstützungen.
Fathers with one child and those with several children receive different levels of state support.

Männer mit drei Köpfen haben oft Schwierigkeiten, einen Job zu finden, um ihre Familie zu unterstützen.
Men with three heads often find it difficult to find a job to support their family.

Der dreiköpfige Familienvater bekommt deutlich weniger Geld.
The three headed family father receives significantly less money.

Eine Hochzeit kann ein prächtiges Ereignis oder eine bescheidene Feierlichkeit sein.
A wedding can be a magnificent event or a modest celebration.

Wenn zwei Adelige heiraten, dann wird die Hochzeit zu einem glamourösen Ereignis.
When two aristocrats marry, the wedding becomes a glamorous event.

Eine königliche Hochzeitsfeier ist grandios und atemberaubenden.
A royal wedding celebration is grandiose and breathtaking.

Experiment 1 Results

	Est.	Est. error	95% CrI
Intercept	−4.05	0.77	−5.61, −2.60
AN1 rating	3.37	0.89	1.68, 5.17
AN2 rating	6.33	0.90	4.61, 8.13
morph. match	−0.06	0.58	−1.21, 1.09
AN1 rating × AN2 rating	−4.01	1.08	−6.16, −1.92
AN1 rating × morph. match	0.14	0.48	−0.80, 1.11
AN2 rating × morph. match	−0.12	0.51	−1.10, 0.88

Experiment 2 Results

	Est.	Est. error	95% CrI
Intercept	−1.95	0.75	−3.42, −0.50
AN1 rating	4.00	0.85	2.32, 5.69
AN2 rating	−1.92	0.82	−3.53, −0.29
morph. match	−0.24	0.65	−1.51, 1.04
AN1 rating × AN2 rating	−1.25	0.97	−3.17, 0.65
AN1 rating × morph. match	0.31	0.56	−0.76, 1.43
AN2 rating × morph. match	0.61	0.57	−0.51, 1.70