Previously, and by that I mean quite some time ago, we set the stage to understand the different types of languages classified by the extent to which certain properties of speech are used to distinguish between lexical terms, with two of the most salient ones being stress and tone. This has led us to two prototypical systems, that of stress languages, and tonal languages, which use stress and tone structures or features respectively at a word level.

But now, we get to the messy part that is everything in between. Stress languages and tonal languages may be thought of as occupying two extremes here, with pitch-accent languages occupying varying places in between, with some adopting more similar characteristics to stress languages, while others adopting more similar characteristics to tonal languages. Therefore, it could be difficult to precisely define what the properties of a pitch-accent language is, since there is not necessarily a one-size-fits-all label one could apply.

As one would have observed previously, stress and tonal languages are examples of typological labels featuring a certain notable property. By definition, stress languages rely on a word-level metrical structure to distinguish lexical terms and other functions, while tonal languages rely on a word-level pitch system to do so. Whether or not a third prototype encompassing pitch-accent languages could exist is still up for debate, although there has been a lot of back-and-forth in how such a prototype may be defined.

The common issue is that pitch-accent languages use a varying combination of properties featured by the two prototypical systems, and defining a prototypical pitch-accent language would in theory, incur overlaps with the other two prototypes. Definition would also be difficult, as the question of what property a certain language must have to be classified as a pitch-accent language, and not a tonal nor stress language.

Donohue’s 2005 publication took some extra liberties in identifying the properties of a pitch-accent language. Rather than going for only the presence of features in a pitch-accent language, Donohue went for a mixture of presence and absence of certain features, listed here:

• Absence of metrical iteration
• Phonologically-specified relative pitch heights
• Contrast on monosyllabic (or monomoraic) words
• Possibility of system to be described with automatic phonological processes and maximally one single diacritic assigned lexically

The last point is a little bit more difficult to grasp, but in essence, it pertains to a possible salient feature of a so-called restricted tone system. Donohue’s other three points mainly pertained to the properties of tonal languages, with the first point rather unusually shown as an absence of a property.

Hyman’s stance is that there is no pitch-accent prototype, and that these systems are part of the continuum between stress and tonal languages, with their own mix-and-match properties of languages that occupy the extremes of this axis. From existing literature at the time of his 2009 publication, he argues that there could have been two different kinds of a pitch-accent language, and that these do not constitute a coherent category. The first category would be where a language inherits properties of both stress and tonal languages, especially when tonal contrasts occur in stressed syllables, while the second category would be a language that does not necessarily meet the requirements or criteria to be classified as a fully tonal language, and so it would be pushed to an ‘intermediate’ category instead.

As such, one suggestion when characterising languages by this sort of prototype is to conduct a tonal analysis or a stress analysis, but never a pitch-accent analysis. While there are certain properties that exist in tonal languages that may exist in the so-called pitch-accent languages, these, according to Hyman, do not provide sufficient support for pitch-accent systems to exist as a separate word-level prosodic prototype.

Therefore, to circle back onto the titular question, a pitch-accent language may just be a language that does not fit nicely in the stress or tonal language prototype. Some linguists propose a continuum between both prototypes, wherein various kinds of pitch-accent languages may fall under, though where exactly a pitch-accent language like Japanese would be classified relative to say, Swedish. Perhaps a better practice is to inquire what roles or functions stress and tone play in the languages which are analysed, averting from a wide umbrella term we call the pitch-accent language, and instead, recognising the intermediate systems that do not neatly fit into the two traditional prototypes.

Now that we have sort of answered the question on how pitch-accent languages are defined or not, let us take a look at some of the more salient examples of such languages.

The case of Japanese

The Japanese language is usually brought up as the textbook example of a pitch-accent language. Although not explicitly written, there are numerous dictionaries that feature this pitch-accent feature, such as the Online Japanese Accent Dictionary. For clarity, the variety of Japanese we will cover here is the one spoken in Tokyo, and is the variety on which Standard Japanese is based on. Hence, Tokyo Japanese is the variety that is the most commonly taught to foreign students as well. Other prefectures may feature varieties that use the same system as Tokyo Japanese, while a more distinct system is used in the varieties spoken in the Kansai region, which is also called the Keihan type. Other regions may not even feature such a system at all.

There are two pitches used in the Tokyo Japanese system, which are high and low (H and L respectively). To get a basic idea on how words in isolation are affected by this word-level prosody, there are certain patterns that are identifiable.

If the first mora (first sound unit in the word) is stressed, then the allocated pitch is H, and drops on the second mora, and would level out. As such, for a word with 2 morae, a pattern could be H-L, and for a word with 3 morae, a pattern could be H-L-L. Conversely, if the last mora is stressed, then the pitch would start off low, and end on a high pitch. For example, words with 3 morae would have the patten L-L-H. If the stress does not occur on the first nor last mora in the word, then the pattern, using a 3-morae word as an example, would be L-H-L.

For words that lack an accent, or stress, then the starting pitch would be low, and rises not to a high pitch, towards the middle of the word. These patterns, while applying to words in isolation, would also cover particles that may be attached to the word. As such, the sequence hashiga (hashi + ga [subject marker]) could mean ‘chopsticks’, ‘bridge’, or ‘edge’ depending on which pattern the sequence follows (H-L-L, L-H-L, L-L-M respectively, M marks middle, a pitch that is in between the L and H pitches).

The case of Swedish

There is no fixed stress in Swedish, and where the stress lands may be used to distinguish between words that otherwise would sound identical. However, stressed syllables may also carry one of two different tones, which are low and high, but more often written as ‘acute’ and ‘grave’, or more mundanely, ‘accent 1’ and ‘accent 2’. These accents vary from dialect to dialect, and so Swedish spoken in Skåne may have different accents from that spoken in Stockholm. However, there are some Swedish dialects where this distinction is minimal or absent, such as Swedish spoken in Finland.

There are a few words (by that I mean at least a hundred of them) that are only distinguished by this very system, like the classic example anden, which translates to ‘the mallard duck’ with the acute accent, and ‘the spirit’ with the grave accent.

Further Reading

Donohue, M. (2005) ‘Tone and the trans New Guinea languages’, Cross-Linguistic Studies of Tonal Phenomena, ILCAA, Tokyo, pp. 33-53.

Hyman, L. M. (2009) ‘How (not) to do phonological typology: the case of pitch-accent’, Linguistic Studies, 31, pp. 213-238.

van der Hulst, H. (2011) ‘Pitch accent systems’, in Volume II. Suprasegmental and Prosodic Phonology, pp. 1-27.

Voorhoeve, J. (1973) ‘Safwa as a restricted tone system’, Studies in African Linguistics, 4, pp. 1-22.