Midterm Practice Problems-F2024
Midterm Practice Problems-F2024
Goad
Fall 2024
Consider the distribution of high tense and lax vowels in the data in Set A from Québec French.
([ʁ] is a uvular fricative, the most common realization of rhotics in Québec French.) If you are a
(native) speaker of Québec French, work only with the data provided – if you overthink the
problem, you may arrive at the wrong answer for some questions.
Set A:
[ʁy], *[ʁʏ] ‘road’ [fɪl] ‘line, lane’ [ʦʏk] ‘toque’
[lynεt], *[lʏnεt] ‘glasses’ [pʊs] ‘grow, push’ [ʦikεt] ‘ticket’
[sɪt], *[sit] ‘site’ [si] ‘if’ [lʏn] ‘moon’
[kʊd], *[kud] ‘elbow’ [suʁi] ‘mouse’ [fu] ‘crazy’
1. Considering only the data in Set A, two possible versions of the rule that accounts for the
distribution of high vowels in Québec French are as follows:
(1) high vowels → lax in closed syllables
(2) high vowels → tense in open syllables
Which version of the rule is more likely to be correct? Support your decision.
Version (1). It is a (near) universal that all languages have tense vowels, but not all
languages have lax vowels. The criterion for determining underlying representations
‘naturalness within segmental inventories’ states that the segments that are underlying are
typically the ones that are most commonly observed across languages.
2. Considering only the data in Set A, is the rule you provided in response to question 1:
a. an allophonic rule c. a neutralizing rule
b. a morphophonemic rule d. we cannot tell from the data provided
4. Taken together, the data in Sets A and B show that the rule applies in both monomorphemic
and multimorphemic words. Recall that variation in the form of a morpheme is called an
alternation (morphophonemic alternation). Do tense and lax high vowels show alternations in
the data in Set B? Clearly explain your answer. (There is room at the top of the next page.)
2
Consider now the additional data in Set C. These data show that the rule provided earlier is
blocked in some contexts.
Set C:
[siv] ‘chive’ [ʒyʒ] ‘judge’ [pʁiz] ‘electrical plug’
[ʁuʒ] ‘red’ [tʁuv] ‘find’ (SG) [myz] ‘muse’
5. Can the contexts in which the rule is blocked be expressed in terms of a natural class? If so,
describe the class in features. Use all the data, not only those in Set C, to answer this question.
[-son, +cont, +vce]
6. The following two-syllable words contain consonant clusters in medial position. Using your
answers to the previous questions, can can you determine where the syllable boundary falls
in these words? If yes, provide the boundary using a period. If not, explain why not.
Three rules are required to capture the changes from Proto-Algonquin to Abenaki. These are
provided below. The third rule is ‘context-free’. Context-free rules apply in all contexts; they are
of the form A → B with no triggering environment. (The term ‘context-free’ is in contrast to
‘context-sensitive’.)
Rule 1: Vowels are deleted word-finally
Rule 2: Singleton obstruents are voiced intervocalically
Rule 3: Vowels are shortened
3
ABENAKI (cont’d)
Your job is to formally capture the changes observed using Optimality Theory. Your overriding
goal is to arrive at a single ranking of constraints that works for all of the Abenaki outputs in the
dataset. Here are the constraints you need:
Faithfulness: MAX: Don’t delete a segment (treat long vowels as two segments, i.e. as VV)
IDENT(VOICE): The input specification for [voice] is preserved in the output
2. Which markedness constraints are dominated (violated in at least some output form) in
Abenaki?
none
In questions 5-6, you must provide tableaux for various input-output pairings. In each case, you
are given the input and candidate outputs to consider. You must provide the ranking and accom-
panying violation marks (e.g. *, *!). You do not need to mark shading in any of your tableaux.
Remember:
• Two violations of a given constraint is less optimal than one violation of that constraint and
should be marked with two asterisks.
• Sometimes the ranking between two constraints cannot be determined; you can put them in
either order in your tableaux, and use a dotted line to separate them.
• In any given tableau, only include the constraints that are relevant for assessing the well-
formedness of the particular candidate outputs in the tableau. You are told how many
constraints you need in each case.
5. Provide a tableau for [oskan] ‘his bone’, putting the constraints you need in the appropriate
order (ranking) for Abenaki. You need three constraints.
Input: /oskani/ NOFINALV *VTV MAX
a. [oskani] *!
F b. [oskan] *
c. [okan] *! **
4
ABENAKI (cont’d)
6. Provide a tableau for [pebon] ‘winter’, putting the constraints you need in the appropriate
order (ranking) for Abenaki. You need all five constraints.
Input: /pepoːni/ NOFINALV NOLONGV *VTV MAX IDENT(VCE)
a. [pepoːni] *! * *
b. [peboːn] *! *
c. [peboni] *! * *
d. [pepon] *! **
F e. [pebon] ** *
Note: Candidate (a) violates each undominated constraint once, so the location of the fatal
violation mark is somewhat arbitrary. By convention, it is the first constraint listed in the
tableau that is marked with ! (i.e., whichever constraint you placed at the top or your ranking).
3. VOWELS IN ROMANCE
The following words are modelled on Romance languages. Provide the rule that accounts for the
alternations observed in the vowels.
/móa/ [múa] ‘grindstone’ /néboa/ [néβoa] ‘fog’
/máo/ [máo] ‘bad’ /meúdo/ [meúðo] ‘little’
/kwéas/ [kwías] ‘which’ /péuɡo/ [péuɣo] ‘open sea’
/beádo/ [βeáðo] ‘deer’ /páus/ [páus] ‘swamp’
/doʁ/ [doʁ] ‘dyer’s weed’ /peóʁ/ [peóʁ] ‘worse’
/ɡémeo/ [ɡémeo] ‘twin’ /saíta/ [saíta] ‘arrow’
/aɡwéo/ [aɡwío] ‘eagle’ /súa/ [súa] ‘only’
/míño/ [míño] ‘mill’ /kóneɡo/ [kóneɣo] ‘canon’
/dóeʁ/ [dúeʁ] ‘to pain’ /maóʁ/ [maóʁ] ‘chief’
1. Provide the rule, in segments, that accounts for the alternations observed.
é, ó → í, ú / __ e, a, o
2. Rewrite the rule from question 1 in features, using standard rule notation. You can include
[±syll] in both the input and environment for your rule, even if this feature is predictable.
You can use the feature [+stress] to indicate that a vowel is stressed, if necessary.
[+syll, +stress] → [+high] / __ [+syll, -high]
Notes: You do not need to include [-low] in the input. Since only the feature [high] is
changing in the rule (→ [+high]), the value for [low] cannot have changed, so /á/ is de
facto excluded from undergoing the rule without having to state the input as [+syll,
+stress, -low]. You could have included the feature [-stress] in the environment. There
are though no adjacent stressed vowels in the data, so you could assume that the absence
of stress on the trigger is due to stressed vowels being targeted. You can tell that stressed
mid vowels (and not mid vowels in general) are targeted by comparing e.g. /ɡémeo/ →
[ɡémeo], *[ɡémio] and /aɡwéo/ → [aɡwío].
5
ROMANCE (cont’d)
4. VOICING IN RUSSIAN
There are many languages where voicing distinctions are phonemic but the contrast between
voiced and voiceless in neutralized in some context. The following forms show that Russian is a
language that fits this profile: some root-final consonants in this language alternate between
voiced and voiceless, depending on their environment.
To account for these data, two phonetically plausible hypotheses are as follows:
Hypothesis A: The alternating root-final consonants are underlying voiced and there is a
rule of word-final devoicing.
Hypothesis B: The alternating root-final consonants are underlying voiceless and there is a
rule of intervocalic voicing.
1. Argue either for Hypothesis A or for Hypothesis B. Discuss how the other hypothesis fails.
Use examples to support your analysis.
Hypothesis A is correct. It is the only hypothesis that can account for all of the data.
Alternating roots like ‘mushroom’ are underlyingly voiced (/ɡrib/) and there is a rule that
devoices obstruents word-finally (/ɡrib/ → [ɡrip]).
Hypothesis B is incorrect. It can only account for half of the data. If alternating forms like
‘mushroom’ were underlyingly voiceless (/ɡrip/) and there were a rule of intervocalic
voicing (/ɡrip-u/ → [ɡribu] ‘mushroom-dative’), we would predict that the non-
alternating forms would also undergo the rule (/xolop-u/ → *[xolobu] ‘bondman-dative’).
6
Consider the following data where voiced stops are derived from voiceless stops in Zoque (stops
includes affricates). The starred forms indicate impossible words in Zoque.
[pata] ‘mat’ [kunu] ‘he fell’ [ŋɡunu], *[ŋkunu] ‘you fell’
[tatah] ‘father’ [kaŋ] ‘jaguar’ [kenba], *[kenpa] ‘he sees’
[win] ‘face’ [ɲanah] ‘his mother’ [liŋba], *[liŋpa] ‘he slashes’
[ʦima] ‘calabash’ [kama] ‘cornfield’ [mjaŋdamu], *[mjaŋtamu] ‘you came’
[nas] ‘earth’ [ʦehʦu] ‘he cut it’ [nʣehʦu], *[nʦehʦu] ‘you cut brush’
[sis] ‘meat’ [ʃohʃahu] ‘they cooked it’
1. Write the rule that derives the allophones from the phonemes in segments.
p, t, ʦ, k → b, d, ʣ, ɡ / n, ŋ __
2. Rewrite your rule from question 1 in features using standard rule notation. There are two
reasonable ways to describe the manner features for the input. Provide both (i.e., provide two
alternative rules that are both consistent with the data).
[-cont] → [+vce] / [+nas] __ [-son] → [+vce] / [+nas] __
3. Having more data on the behaviour of a particular class of segments could have helped you
decide which rule formulation in your answer to question 2 is correct. Which class of
segments is this? Explain your answer, making reference to the features in the two alternative
rules that you provided.
The necessary class of segments is fricatives. We only see voiceless fricatives in the data
and we have no data where fricatives occur after nasals so we don’t know if they undergo
voicing in this context. If they do undergo voicing, then the input to the rule would need
to be [-son]; if they do not undergo voicing, then the input would need to be [-cont].
4. Return to the way your rule is expressed in question 1. Do you predict from this formulation
that the stops in these new data will surface as voiced or voiceless in the way that they do?
Explain why or why not.
No. The rule in my answer to question 1 is simply a list of segments. As expressed, this
list is arbitrary, as natural classes are not captured through lists of segments.
5. From the way your rule is expressed in question 2, do you predict that the stops in these new
data will surface as voiced or voiceless in the way that they do? Explain why or why not.
Yes. Expressing the rule in features in my answer to question 2 predicts that all voiceless
stops (or obstruents, as per the second formulation of the rule) will be targeted by the
rule, even those that are not included in the list in my answer to question 1.
7
Examine the following data from Oʼodham, an Uto-Aztecan language spoken in Arizona and
Sonora (Mexico).
[ʤud̪um] ‘bear’ [t̪ aapn̪aɲ] ‘split!’ [ʧɨhaɲ] ‘hire’
[ʂɨɨʔɨ] ‘wolf’ [siʔaliɡ] ‘dawn’ [biʤ] ‘turn around’
[ʔɨlid̪aɡ] ‘skin/bark’ [n̪aat̪ o] ‘finish’ [mulin̪] ‘break’
[ʂoɲikon̪] ‘beat’ [siakam] ‘hero’ [ɲuukud̪] ‘care for’
1. Assume that all segments that appear in these data are phonemic in Oʼodham and that
segments that are not present in the data do not exist in the language. Based on these data, fill
in the chart below with the consonant phonemes that occur in Oʼodham; place voiceless
segments to the left of their voiced segments. Ignore vowels (for now).
labial dental alveolar retroflex pal-alv palatal velar laryngeal
stop p b t̪ d̪ k ɡ ʔ
affricate ʧ ʤ
fricative s ʂ h
nasal m n̪ ɲ
lateral l
2. In this question, we will focus on place features for coronal consonants. As shown in the
table below, [CORONAL] is specified for all coronal consonants in Oʼodham. Fill in ALL AND
ONLY the additional place features that are needed to capture place contrasts among coronals
in this language. You may be provided with more lines than you need. (Keep in mind that
laryngeal features (for stops/affricates) and manner features (for stops/affricates, fricatives,
nasals, laterals) will also be specified in these segments, ensuring that each segment can be
uniquely identified. The feature [strident], though, plays no role in this language.)
t̪ ʧ s ʂ n̪ l
[CORONAL] ● ● ● ● ● ●
[anterior] + – + –
Note: You do not need minor place features under [CORONAL] for the nasal or lateral because
there are no contrasts among coronal nasals or laterals in Oʼodham. For example, this
language does not contrast /n̪/ and /ɳ/. Thus, as soon as /n̪/ has been labelled as [COR] (in
combination with [+nas]), the segment has been uniquely identified.
8
OʼODHAM (cont'd)
3. In this question, we will focus on place features for dorsal consonants. As shown in the table
below, [DORSAL] is specified for all dorsal consonants in Oʼodham. Fill in ALL AND ONLY the
additional place features that are needed to capture place contrasts among dorsal consonants
in this language. You may be provided with more lines than you need. (Keep in mind that
manner features will also be specified in these segments, as noted above.)
k ɲ
[DORSAL] ● ●
no Fs needed
4. Consider now the [+high] vowels in Oʼodham (you can ignore all other vowels). Which
articulators and which features are needed to contrast the three high vowels from each other?
Fill in the table below with ALL AND ONLY the articulators and features that are needed. (You
can leave out [high].)
i ɨ u
[DORSAL] ● ● ●
[back] - + +
[LABIAL] ●
In the data below (modified), [i], [iː], [ɛ], and [ɛː] are all in complementary distribution. The
phoneme is /iː/. Treat long vowels as separate allophones from short vowels. This problem is a
little harder than what you will find on the midterm but it is good practice for derivations.
[ʦit'] ‘there is’ [χɛːq] ‘scratch’
[q'ɛt'] ‘be morning’ [ʧiːtɬ] ‘top’
[t'ɛːq] ‘cold’ [q'ɛːχɛʔ] ‘become black’
[χɛp'] ‘get touched’ [qliːm] ‘weak’
[ʦiːx] ‘reach’ [sq'ɛʔ] ‘slices of dried salmon’
[ʦip'] ‘feel cold’ [χɛʔ] ‘be lost’
[k'iːn] ‘few’ [niʔʧ] ‘high seas’
[tmiːx] ‘earth, ground’ [kʦiʔʦ] ‘person with magic power’
[nəq'ɛːɣos] ‘wise’ [χɛːʁʔ] ‘come out’
[xɑʁɛʔn] ‘white man’ [ʔiː] ‘be here’
[tɬɛːq] ‘arrive’ [mɛːχætɬ] ‘black bear’
9
SQUAMISH (cont’d)
1. To obtain the correct distribution of allophones for /iː/, three rules are required. (There is no
typo in Rule C: the input truly is a short vowel.) For each rule, describe the segments that
trigger the rule, in words or features, taking account of all natural classes:
Triggers:
Rule A: iː → i / __ p', t', ʔ Describe in features: [+CG]
Rule B: iː → ɛː / __ q, χ, ʁ Describe in words: uvulars
Rule C: i → ɛ / q, q', χ, ʁ __ Describe in words: uvulars
2. Some of these rules need to apply in a particular order to obtain the correct outputs. If two
rules need to be ordered, the output of the first rule will be the input to the second. Consider
Rules A and C for the form /q'iːt'/ → [q'ɛt'] ‘be morning’. Do these rules need to apply in a
particular order? Provide two derivations (where the order differs) to support your answer.
3. Consider Rules B and C for the form /tɬ'iːq/ → [tɬ'ɛːq] ‘arrive’. Do these rules need to apply
in a particular order? Provide two derivations (where the order differs) to support your answer.
Only Rule B affects this form, regardless of ordering, so it doesn’t matter what the
ordering is relative to Rule A:
Correct: /tɬ'iːq/ ‘arrive’ Correct: /tɬ'iːq/ ‘arrive’
Rule B: tɬ'ɛːq Rule C: –
Rule C: – Rule B: tɬ'ɛːq
[tɬ'ɛːq] [tɬ'ɛːq]
4. Consider Rules A and B for the form /ʦiːt'/ → [ʦit'] ‘there is’. Do these rules need to apply in
a particular order? Provide two derivations (where the order differs) to support your answer.
Only Rule A affects this form, regardless of ordering, so it doesn’t matter what the
ordering is relative to Rule B:
Correct: /ʦiːt'/ ‘there is’ Correct: /ʦiːt'/ ‘there is’
Rule B: – Rule A: ʦit'
Rule A: ʦit' Rule B: –
[ʦit'] [ʦit']
5. If you had a form like [t_q'] where the line represents one of the four allophones under
discussion, could you predict which vowel would surface? Refer to the classes you used in
question 1 to explain your answer.
We cannot predict since the following consonant is both [+CG] and uvular. The vowel
could be [i] since this consonant is [+CG] (as per Rule A) or it could be [ɛː] since the
following consonant is uvular (as per Rule B).