• Keine Ergebnisse gefunden

The categorisation of derivational suffixes and accent assignment

3. Derivational morphology and accent assignment

3.1. Stressed and unstressed prefixes

3.2.5. The categorisation of derivational suffixes and accent assignment

To conclude our examination of derivational suffixes and accent assignment, we give a characterization of each suffix, taking complex derivations into account. However, before we do this, let us first consider the following data.

(59) Complex derivations and accent assignment

Suffix Accent 1 Accent 2 Gloss

Suffix Accent 1 Accent 2 Gloss

For each suffix, we attempted to examine complex derivations with unstressed prefixes, stressed prefixes from a verbal source, and stressed prefixes from a nominal source. That is, our intentions were to find out how these suffixes combine with lexical Accent-1 prefixes and stressed unspecified prefixes. This task proved to be quite challenging, since the actual morphological make-up of a complex derivation is not always transparent. However, in most cases the accent of the entire derivation was able to throw light on the derivational make-up. In the following we consider each suffix individually. We start out with the (relatively) clear cases of suffixes in which the categorisation corresponds exactly to the predicted scenario.

Finally, we discuss the less clear-cut cases, namely, {-ig} and {-lig}.

Most of the suffixes that fit into category (55) (i), as we would predict, never affect the accent of the complex derivation that they are in. We can say this confidently for {-else}, {-er}, {-(i)sk}, {-løs}. The suffix {-dom} does not appear in any complex derivations and therefore remains in this unspecified category for lack of any evidence to the contrary.

Although its history suggests that it possibly belongs to the same category as {-skap}, i.e., scenario (55) (v), we do not have enough evidence to affirm this.

When {-bar} or any of the suffixes in scenario (55) (i) are found in complex derivations with unstressed prefixes, e.g. be'regnbar1, the lexical specification of the unstressed prefix, in this case /be-+/, wins out and Accent 1 is always the outcome. The stressed prefixes are not as lucid. Often the lexical specification of the prefix prevails, e.g. 'anvendbar1 ‘useable’

(< 'anvende1), yet other cases seem to defy the lexical specification. In, for example, the derivation 'utnyttbar2 ‘exploitable’, the question arises whether its internal composition comes from 'utnytte1 + {-bar}, or from {ut-} + 'nyttbar2. The accent tells us the latter; the semantics would choose the former. Bracketing paradoxes are a known phenomenon in the

morphology-phonology interface (cf. Kiparsky 1982), and the accent seems to be telling us the phonological story here. Stressed prefixes that have two allophones, one lexically specified when forming a verb, and one unspecified when forming anything else, i.e., noun or adjective, often seem to be involved in bracketing paradoxes. We believe, however, that in cases where the semantics seem to suggest that the bracketing should be for example [utnytte1]VERB + {-bar}, but the accent tells us that it should be three morphemes {ut} + {nytte} + {-bar}, all added on the same level, that the accent is indeed telling the phonological facts. We will return to this topic shortly when considering the behaviour of {-sel} and complex derivations. There are two more suffixes, which belong in the unspecified category: {-ing} and {-som}, that have a few peculiarities, therefore, we discuss them separately.

First, we consider the suffix {-ing}, which is plainly unspecified. It appears not to have any effect on accent except when it adds a syllable to monosyllabic stems, resulting in Accent 2 as we saw in (57g). However, sometimes {-ing} can be found with differing stress patterns (59g).

In these complex derivations, we often find the stress on the prefix or the base (e.g. 'opplyse1

‘to illuminate’, 'opplysning1/opp'lysning1 ‘illumination’). Kristoffersen (p.c.) tells us that often the difference in stress can also signify a difference in meaning. He gives the example of 'avdele1 ‘to divide’, which has the more semantically transparent derivation 'avdeling1 ‘the act of separating something’, and av'deling1 that means more ‘division’ as in the division of a company. Although this divergent stress patterns would have to be accounted for in the rules of Standard East Norwegian stress, we will put that aside here since our analysis is concentrating on lexical accent assignment.

Stress is also an issue with the suffix {-som}. In complex forms containing {-som}, stress is usually found on the verb stem, and seldom on the prefix ('oppfinne1 ‘invent’, opp'finnsom1

‘inventive’). This is so even when there is no prefixed verbal base (e.g. opp'merksom1

‘observant’) – unless the prefix is a nominal prefix, and thus unspecified ('tiltaksom2

‘enterprising’, 'tiltak2 ‘enterprise’) which remains default Accent 2. However, we know that nominal prefixes behave differently from verbal prefixes in Germanic languages.56 Thus, the fact that stress also differs in complex derivations is not disturbing to our analysis. Once again

56. For example we know in English that nominal derivational suffixes attach at a different lexical level (Level 2) than verbal suffixes (Level 1) (Kiparsky 1982). This can be illustrated by the resyllabification that applies in damnation, where the /n/ makes up the onset of the second syllable, which does not apply to damning.

we believe loans play an important role, since Norwegian has modelled many of these complex derivations on German (opp'merksom1 ‘observant’). Unfortunately, more investigation into the effect of stress and accent in such complex derivations would require a solid conception of how stress assignment works in Standard East Norwegian, how it functioned in MLG and High German, and the interaction between the two stress systems under borrowing – which is unfortunately not possible within the framework of this thesis. Therefore, since the accent is in any case never affected by {-ing} and {-som}, we categorise them with the unspecified suffixes.

The suffix {-sel} is our first and only candidate for belonging to scenario (55) (ii), i.e., only counts as a full syllable on the surface after accent assignment and epenthesis. Once again, the synchronic phonological facts governs the analysis. This nominalizing suffix is native, stemming from the ON suffix {-sl} or {-sla}. When suffixed to a monosyllabic word in (57m), we saw that the derivation retains Accent 1. This could indicate that {-sel} is either lexically specified for bearing Accent 1, or that it does not affect the accent of a monosyllabic stem at all. That is, accent assignment proceeds as if no additional syllable has been added. This is in fact our analysis. We claim that {-sel} is unspecified for tone, and underlyingly non-syllabic /-sl/. It receives an epenthetic vowel post-lexically after accent assignment. Historically this was indeed the case, and synchronically its behaviour shows that it is non-syllabic.

Unfortunately, as we mentioned earlier, {-sel} only takes monosyllabic verb stems as a base. Therefore, there are no {-sel} derivations with monomorphemic polysyllabic stems to test whether {-sel} does, or does not carry lexical accent. However, complex derivations consisting of more than one affix can. As we see in (59m), the derivation 'påkjørsel ‘crash’ finally presents us with a word suffixed with {-sel} that contains a disyllabic trochee. We now have 'σ σ + {-sel} and get Accent 2, verifying our assumption that {-sel} is not lexically specified for Accent 1.

We posit that when {-sel} attaches to monosyllabic bases, the only possible outcome is Accent 1, because accent assignment only has a monosyllabic domain ('σ + /sl/) to work with.

Epenthesis happens post-lexically, after accent assignment. In 'påkjørsel ‘crash’ there is no lexical specification, yet two full syllables, thus the result is Accent 2. Here, it appears as if we have a case of a bracketing paradox, since the correct internal organisation according to the semantics would be 'påkjøre1 ‘to run into’ + {-sl} with a lexically specified {p+å}. This

derivation would have a lexically specified {p+å} because we are dealing with a verb (cf.

section 3.1.2). Yet, this lexically specified verbal prefix would incorrectly give us lexical Accent 1. Thus, the Accent 2 of 'påkjørsel2 tells us that we are not dealing with a verbal derivation. The phonology tells us that the organisation must be {på} + {kjør} + {sl}, since the result is a noun with unspecified Accent 2. The allomorph {på} is definitely unspecified here.

We claim it is unspecified, because we are forming a noun. If we were dealing with a real verbal derivation, such as 'påskjønnelse1/'påskjønning1 ‘acknowledgement’ from the verb 'påskjønne1 ‘to acknowledge’, then we indeed have [påskjønn]1VERB + {-else}/{-ing} with the lexically specified {p+å} allomorph. The accent and semantics confirm that we are deriving a noun from a verb. In 'påkjørsel, we are forming a noun with three morphemes {på} {kjør}

{sl}, none of which are lexically specified, since we are constructing a noun, and the result is Accent 2. This, of course, also follows for {ut-} and all other particles, e.g. 'utkjørsel2

‘driveway’, 'anførsel2 ‘command’, 'avkjørsel2 ‘exit’.

Concerning scenario (55) (iii), the only lexically specified suffix out of our sample is {-+isk}.

As we mentioned in section 3.2.2, we claim that there are two {-(i)sk} suffixes. One that is specified, and one that is not specified. It is often difficult to determine which /-(i)sk/ we are dealing with when lexical accent is already present in the stem. However, our verdict holds that there are two allomorphs: One lexically specified {-+isk} and one unspecified {-(i)sk}.

The only clitic that we can add to scenario (55) (iv) is possibly {=het}, which behaves very much like the definite articles. Clitics are added after accent assignment. Therefore, whatever the accent of the preceding form is in isolation, this will also be the accent of the entire form that includes {=het}. Accent assignment is performed and then {=het} is attached.

Scenario (55) (v) only includes the suffix {-skap} that makes up a prosodic word by itself. It acts no differently with complex derivations containing another affix + {-skap}ω. The morpheme {-skap}ω acts just like a second member of a compound, i.e., the first member determines the accent of the entire compound, as we will see in Chapter 4.

The remaining two suffixes, {-ig} and {-lig}, which have been categorised as unspecified until now, are very puzzling, as can be seen in the following:

(60) Complex derivations and accent assignment with {-ig}{-lig}

Suffix Base Accent 1 Accent 2 Gloss

be'standig1 continuous

'andakt2 > an'dektig1 devout

'forskjell2 > for'skjellig1 different a. {-ig}

'avhenge1 > 'avhengig2 dependent

b. {-lig} be'vislig1 natural

'oppbygge1 opp'byggelig1 uplifting

'tilråde1 til'rådelig1 advisable

'he+der1 'hederlig2 honest

The suffixes {-ig} and {-lig} are found in complex derivations where stress is on the stem vowel, which sometimes also undergoes umlauting (e.g. 'andakt2 ‘devotions’ > an'dektig1

‘devout’), and accent change. Most forms keep their lexical accent as in 'oppbygge1 ‘to build up/to uplift’ opp'byggelig1 ‘uplifting’, however, some complex derivations with {-ig} do not (e.g. 'avhenge1 ‘it depends’, 'avhengig2 ‘dependent’). But what is most difficult to explain is the fact that {-ig} and {-lig} are sometimes blind to lexical accent in derivations such as, 'a+delig2

‘nobel’ and 'ri+dderlig2 ‘knightly’. These stems both have Accent 1 in isolation and when suffixed with another prosodic word like {-skap} e.g. 'adelskap1 ‘nobility’, 'ridderskap1

‘knighthood’. As we illustrated with the nomen agentis {-er}, a good deal of these oddities we believe are connected with loanword incorporation. For example, a word like 'andakt2

‘devotions’ was borrowed from German ('andacht) through Danish, and fit in with the other native nouns with stressed prefixes that recieves Accent 2. The word an'dektig ‘devout’, however, according to the Bokmålordboka and the Nudansk Ordbok, was derived from 'andakt2

and was not borrowed from German. Thus, as far as our approach is concerned, it should have Accent 2. However, German also has the form 'andächtig with umlaut. These Germanic words are so similar it is difficult to deduce the true origins.

As to the insensitivity to lexical accent, perhaps the key lies in the second syllable of the stems. Schwa is often invisible to stress, and perhaps we are dealing with lexically specified monosyllabic stems as far as {-ig} and {-lig} are concerned. Monosyllables have a special status as we shall see in Chapter 5. Concerning {-ig} and {-lig} showing different stress patterns with prefixed forms, as already noted above with the suffixes {-ing} and {-som}, we

will not embark upon the investigation of the effect of stress and accent in complex derivations in this thesis but leave it for future research.

Given that these suffixes otherwise do not appear to affect accent, we classify {-ig} and {-lig} as unspecified under scenario (55) (i), because of their behaviour with monosyllabic and polysyllabic stems, as we saw in (57) and (58). We add to this the annotation that lexically specified disyllabic words containing a schwa in the second syllable are monosyllabic for {-ig}

and {-lig}. Thus, lexical specification will not surface in derivations with these words + {-ig}/

{-lig} just as it does not surface in lexically specified monosyllabic stems + {-ig}/{-lig}.