In a Bantu language, a vowel is deleted from a word, but the high tone that was associated with that vowel appears on the adjacent vowel. In autosegmental phonology, what does this 'floating tone' phenomenon demonstrate?
AThe deletion rule is incomplete — the vowel was reduced rather than fully deleted
BTones exist on an independent tier; when the segmental tier loses the vowel, the tone remains on the tonal tier and reassociates to an adjacent segment
CThe rule deleting the vowel also includes a compensatory copying step that transfers tone to the neighbor
DHigh tones spread automatically across all adjacent segments regardless of vowel deletion
This is one of the core empirical motivations for autosegmental theory. A linear representation cannot explain tone survival after vowel deletion — if tones are properties of segments, a deleted segment takes its tone with it. The autosegmental solution is that tones are on an independent tier connected to segments by association lines. When the vowel is deleted on the segmental tier, the tone 'floats' on the tonal tier and then reassociates with an adjacent vowel under well-formedness constraints. The feature outlives the segment because it was never part of the segment — it was a separate object linked to it.
Question 2 Multiple Choice
Finnish vowel harmony requires all vowels in a word to share the same value for [back] — all back or all front. How does autosegmental phonology explain this more parsimoniously than a linear approach?
AA single copying rule iterates left-to-right, applying the [back] value of the first vowel to each subsequent vowel
BA word-level filter rejects any word containing both [+back] and [-back] vowels after derivation
CA single [±back] specification on the vocalic tier associates with all vowels in the word, so they share one feature rather than each having their own
DConsonants between vowels block the spread of [back], so harmony operates only within syllables
The autosegmental account replaces multiple identical feature specifications with a single one. In a linear model, if four vowels all have [+back], you need four separate [+back] features and rules that copy or agree. Autosegmentally, there is one [+back] token on the vocalic tier that simply associates with multiple vowel slots — harmony is just a single feature with a wide scope of association. This is both more economical and naturally handles why harmony ignores consonants (consonants are on the segmental tier; the vocalic [back] tier skips over them). It also immediately explains why adding a new vowel to a word is automatically harmonized.
Question 3 True / False
In autosegmental theory, a single tonal specification such as [H] can be associated with multiple vowel positions simultaneously, producing a level high tone across an entire domain.
TTrue
FFalse
Answer: True
This is a defining feature of autosegmental representations. Association lines connect one tier to another, and a single tonal element can link to multiple segments. This is how spreading is represented: not as successive copying but as broadening the domain of association. A [H] associated with three vowels produces a level high tone across all three; if it were later restricted to one vowel, the other associations would be severed. This multilinear association is what makes autosegmental representations fundamentally different from feature matrices in standard segmental phonology.
Question 4 True / False
The Obligatory Contour Principle (OCP) prohibits identical adjacent features from appearing anywhere on any tier, requiring that most adjacent segments differ in at least one feature.
TTrue
FFalse
Answer: False
The OCP is a constraint on a *single* tier that prohibits two identical adjacent specifications on that tier. It applies tier-specifically, not as a general requirement for segments to differ in all features. The OCP on the tonal tier says you cannot have [H][H] adjacently — they must be merged into one [H] with dual association. But it says nothing about, say, adjacent consonants sharing the same place feature on the segmental tier. It is also not an absolute inviolable constraint but a strong tendency that interacts with other requirements. The formulation 'all adjacent segments must differ in at least one feature' is both too strong and confused about what the OCP governs.
Question 5 Short Answer
A geminate consonant (like the double-t in Italian 'notte') resists deletion processes that would remove a 'single consonant.' How does autosegmental representation explain this resistance?
Think about your answer, then reveal below.
Model answer: In autosegmental phonology, a geminate is represented as a single consonant specification on the segmental tier associated with two timing slots (two positions on the skeletal or timing tier). A deletion rule that removes 'one consonant' removes one timing slot. But since the single consonantal specification still has an association to the remaining timing slot, the consonant is not gone — it surfaces as a regular short consonant. Deleting a true geminate requires severing all associations, which is a more costly operation than single-segment deletion. The geminate's 'weight' comes from this dual timing association, not from being two separate consonants.
This is the elegance of autosegmental geminates: they explain why geminates behave as phonologically heavier (they occupy two timing slots) and why they resist deletion (deleting one association still leaves the consonant present). Linear phonology would need two separate identical consonant segments to represent a geminate and an ad hoc constraint preventing both from being deleted simultaneously.