Refining our analysis

So far, we have seen only cases in which a final bilabial, alveolar, or velar stop assimilates to a following stop in Mallorquí and Central Catalan. We accounted for the role of stops by specifying the features [-sonorant, -continuant] in the structural description of both the target, C1, and trigger, C2, of assimilation in rules (10) and (11).

Remember that the target of a phonological rule of assimilation is the sound that undergoes a change, while the trigger is the sound that "spreads" its features to the target.

The formal rule statements we proposed earlier make predictions that need to be investigated before we can be satisfied that our analysis is correct. We specified the features [-sonorant, -continuant] in our rules because in all of the cases we have seen so far, the targets and triggers of gemination have been oral stops, which are characterized by these features. One argument we need to make in order to confirm our analysis, is that restricting the sound class(es) participating in the rule to oral stops is not too restrictive. Oral stops are [-sonorant] segments, or obstruents, but they are not the only obstruents: fricatives and affricates also belong to this class.

Fricatives and affricates differ from stops in having the feature [+continuant]. Fricatives are straightforwardly [-sonorant, +continuant]. The standard view of affricates is that they should be represented as contour segments for the feature [continuant], as shown in diagram (12). On this view, an affricate is doubly specified for [continuant], with an initial [-continuant] closure phase, followed by a [+continuant] phase, which accounts for the fricative-like release.

(12) / ʧ /
  [-continuant] [+continuant]

Moving forward, the question to ask is whether any of the [+continuant] obstruents are affected by place assimilation (as targets), or alternatively, provide its conditioning environment (as triggers). If so, we would need to modify the features we used in our rules. Before settling on an analysis that restricts gemination to stops, then, we need to make the case that neither fricatives nor affricates participate in the process. Here are some possible outcomes, and what they would mean for our analysis.

  • The rules we have already written make a prediction that we have not discussed. We have defined our target class as [-sonorant, -continuant]. If (as we believe) affricates are specified as [-continuant, +continuant], our rule as it is now predicts that affricates as well as stops should be targets. Therefore, if further data show that affricates but not fricatives are targets, then we will have been correct in limiting the target class to [-continuant] obstruents, and no change in our definition of the target class will be necessary.
  • If it turns out that both fricatives and affricates are targets in addition to stops, then we will need to broaden the structural description of the target class by dropping [-continuant] from our rules.
  • The third scenario is a little trickier. If neither fricatives nor affricates are targets, then our assumption that only stops can be targets will be correct. In this case, however, it won't be enough to have defined targets as [-sonorant, -continuant] in our rules. To exclude affricates, we will need to add a negative condition stating that [+continuant] obstruents do not undergo the rule.
  • Given the system of phonological features we are using, there is one scenario we do not expect: we should not find that stops and fricatives, but not affricates, pattern as targets. This is because given the way the obstruent subclasses are specified for the feature [continuant], there is no way to group stops and fricatives as a class without the affricates.

Once we have determined which of these scenarios applies to the sound class that is targeted by the rule, we will need to do the same for the conditioning (trigger) class.

Turning first to the targets of assimilation, the examples in (13) show that fricatives do not undergo assimilation in either Central Catalan or in Mallorquí.

(13) Fricatives are not targets in either Mallorquí or Central Catalan.
Central Catalan (1st button) and Mallorquín (2nd button)
tres pares [tɾəs paɾəs] *[tɾəf paɾəs] 'three fathers'
tres caras [tɾəs kaɾəs] *[tɾəʃ kaɾəs]
*[tɾəx kaɾəs]
'three faces'
tres bocas [tɾəz bokəs] *[tɾəv bokəs] 'three mouths'
mateix dia [məteʒ diə] *[mətez diə] 'same day'
mateix peu [məteʃ pɛw] *[mətef pɛw] 'same foot'

The examples in (13) tell us that we are correct to include [-continuant] in the description of the target sound class. Moving on, the examples in (14) show that affricates are also unaffected by the rule.

(14) Affricates do not assimilate in either Mallorquí or Central Catalan.
Central Catalan (1st button) and Mallorquín (2nd button)
puig gros [puʤ gɾɔs] *[pug gɾɔs] 'big hill'
puig bo [puʤ bɔ] *[pub bɔ] 'good hill'
puig contigu [puʧ kuntiγu] *[puk kuntiγu] 'contiguous hill'
puig tedios [puʧ təðius] *[put təðius] 'tedious hill'

The phonetic transcriptions given are for Central Catalan. Note that in the Mallorquín word contigu, the first vowel is [o], not [u] as it is in Central Catalan. This is because processes of vowel reduction work differently in the two dialects.

Now we know that the targets of obstruent place assimilation really are stops; fricatives and affricates are not affected by the process. We were correct in stating that target sounds must [-continuant], but we will need to add to the final version of our rules a statement that targets may not be [+continuant].

We are now ready to have a closer look at triggers - the sounds that provide the conditioning environment for the assimilation rule, and whose place features are projected onto targets.

Currently, we have no data bearing on the behaviour of affricates in trigger position. When we look at fricatives, however, we find an interesting difference between Central Catalan and Mallorquí. Recall that only the alveolar stops /t, d/ assimilate to a following C in Central Catalan. In order to show whether or not a fricative can trigger place assimilation, we need to see cases in which /t, d/ precede a fricative with a different place of articulation. The examples in (15) show that fricatives do not trigger place assimilation in Central Catalan. (Note that we are using [pf] for a voiceless labiodental stop. By superscripting the 'f' we intend to show that this is a stop, not the affricate [pf]. The IPA does not have a standard symbol for a labiodental stop.)

(15) Fricatives do not trigger assimilation in Central Catalan.
Critical example for Central Catalan
plat felic [plapf fəlis] *[plat fəlis] 'happy plate'

Thus, we see that in Central Catalan, the second consonant in the target cluster may not be [+continuant]: neither affricates nor fricatives, characterized by [+continuant], trigger assimilation. However, Mallorquí differs from Central Catalan in that fricatives do trigger assimilation. In the Mallorquí examples in (16), /p/ in cap and /k/ in puc are alveolar [t] before [s], and /t/ in plat surfaces as [pf] before [f], just as in Central Catalan.

(16) Fricatives trigger assimilation in Mallorquí.
Critical examples for Mallorquí
cap sentit [kap səntit] *[kat səntit] 'no sense'
puc sentir [puk sənti] *[put sənti] 'I can feel'
plat felic [plat felis] *[plat fəlis] 'happy plate'

Based on the examples in (16), we can see that restricting the trigger class to [-continuant] in Mallorquí is too narrow, since fricatives also trigger place assimilation. In Mallorquin, therefore, we need only the feature [-sonorant] to define the trigger class. In Central Catalan, the trigger class is better defined by the features [-sonorant, -continuant].

Note that the feature sets we are associating with the trigger class in Central Catalan and Mallorquí admit affricates in both dialects. If further data should indicate that affricates don't trigger place assimilation in Central Catalan, we would impose a negative condition excluding [+continuant] obstruents, as we did for the target class in both dialects. However, if we should find that affricates don't trigger place assimilation in Mallorquí, then we would have a real problem. See if you can work out why.

What we have learned from the examples we have now seen is that stops are the targets of geminating place assimilation, or the class [-sonorant, -continuant]. It isn't quite enough just to say that targets may be [-continuant] since affricates, which are both [-consonant] and [+continuant], cannot be targets. Therefore, we need a "tag" on our rule that specifically excludes [+continuant]. These conditions on the tgarget class are the same in both Central Catalan and Mallorquí. Conditions on the trigger class are a little different in the two dialects. In Central Catalan, only stops are triggers. However in Mallorquí, both stops and fricatives are triggers. In Catalan, then, the "not [+continuant]" tag is required not only for targets, but for triggers as well. For Mallorquí, the trigger class can be defined as [-sonorant] more broadly, without specifying a value for the feature [continuant].

Near-final versions of our place assimilation rules for stop assimilation in Central Catalan and in Mallorquí are given in (18) and (19).

(18) Central Catalan alveolar stop assimilation (revised)
[-sonorant, -continuant, +coronal, +anterior]
[-sonorant, -continuant, αplace]
Condition: [+continuant] obstruents do not participate in this rule either as targets or triggers.

(19) Mallorquin stop assimilation (revised)
[-sonorant, -continuant]
[-sonorant, αplace]
Condition: [+continuant] obstruents do not undergo this rule as targets.

A note on things to come: We are not quite finished this discussion of geminating place assimilation; this is why I called the rules in (18) and (19) "near-final" versions. Here's a question for you to ponder: I think we've established the role of the feature [continuant] in defining the target and trigger classes for the rules for these dialects. One loose end is that we don't have examples showing whether affricates can be triggers in either language. The analysis as we have it so far predicts that they might or might be in Central Catalan, and it predicts that they must be able to be triggers in Mallorquí. We'll be looking for data that will help us to answer this question. Another thing to ponder: have we really proven that targets must be [-sonorant] in this sound process, or have we just taken this for granted? Retrace the steps in our argument to see. What might be missing from our analysis? (Hint: what sonorant sounds are also [-continuant]?) Check back in November for further discussion!