Trees and Waves

Historical Linguistics and Language Families

Course 1 ยท Chapter 6 ยท Trees and Waves

Chapter 5 drew Indo-European as a tree. A tree is a good picture of one thing: a language splits into daughters and they drift apart. But languages also meet. Words are borrowed, features spread between neighbours, and unrelated languages can end up looking alike. This chapter looks at what the tree leaves out.

The idea in a box
The tree model assumes languages change only through social splitting and divergence. The wave model pictures an innovation that starts in one area and spreads speaker to speaker and dialect to dialect, like waves on water. A full account of language history needs both.

The Tree and Its Problem

Wikipedia's article on the wave model says the tree model "struggles to explain some shared features, especially in the Germanic languages". A tree can show that two branches share a recent ancestor; it cannot easily show that a change spread across the boundary between branches after they had separated.

ModelPictureGood for
Treea line that forks and never rejoinsdescent, subgroups (Chapters 3 to 5)
Wavean innovation spreading outward, with waves overlapping in intersecting patternsdialect continua, areal phenomena, features shared across branches

How the wave model works

  • An innovation starts in one area and spreads speaker to speaker and dialect to dialect. Innovations can overlap in intersecting patterns, which the tree cannot easily represent.
  • Dialectology uses linguistic atlases and maps to show isoglosses (the boundaries of a feature) and dialect boundaries, including fuzzy ones such as the Croissant and the Rhenish fan (the page's own examples).
  • Schmidt's continuum metaphor: a smooth slope becomes a stepped line, with each step a dialect. Some steps fade, while others, like Standard German, can spread and replace local dialects.
  • Advocacy for the model is credited to Johannes Schmidt and Hugo Schuchardt.
Watch out: it is not tree versus wave
The wave model has "contributed greatly to improve, but not supersede" the tree model. Some scholars argue it should replace the tree for genealogy, and the page says linkages (families descended from a former continuum) must be analysed with waves. It was little accepted in the 20th century except for dialect continua and areal phenomena, and has regained popularity because of the tree's shortcomings. Use the tree for descent and the waves for spread.

Language Contact

Wikipedia's article on language contact lists what can happen when speakers of different languages meet:

OutcomeWhat it isExample from the source
Borrowing"The most common way that languages influence each other is the exchange of words"; influence can reach morphology and grammarNewar absorbed noun inflection and verb tenses from neighbouring Indo-Iranian languages
Code-switchingmultilingual speakers use multiple languages in a single conversation(none given)
Pidgin, creolea pidgin forms among people without a common language and may become a creole; some linguists deny a creole must pass through a pidgin stageNdyuka (Aukan) and Saramaccan, in Suriname, draw their vocabulary mainly from Portuguese, English and Dutch
Mixed languageforms among communities fluent in both source languages; keeps much of the parents' grammatical and phonological complexity(none given)
Language shift and deathone language replaces another, most often the one with higher prestige; can lead to endangerment or extinction(none given)

Loanword cases

  • Latin and French into English: the page describes the large-scale import of words from these languages in the 16th and 17th centuries as more significant than modern English borrowing into other languages.
  • Iranian into Armenian: Armenian borrowed so many Iranian words that it was at first considered a divergent Indo-Iranian branch. This is a warning for Chapter 3: heavy borrowing can fake relationship, so correspondence sets must come from basic vocabulary. Armenian is, in the Indo-European table of Chapter 5, a branch of its own.
  • English into Hindi and Scots, Afrikaans into South African English, and Chinese into Japanese, where some Japanese coinages based on Chinese forms were later reborrowed into Chinese.

During much of the Middle Ages, upper-class English speech was "dramatically influenced" by Norman, the page says, to the point that it often resembled a dialect. It also gives English phrases adapted from French where the adjective follows the noun: court-martial, attorney-general, Lake Superior.

Linguistic Areas (Sprachbunds)

A sprachbund (linguistic area) is a group of languages that share "areal features" from geographic proximity and language contact. The languages may be genetically unrelated or only distantly related. A sprachbund is defined by features shared for reasons other than common descent. Without the history of a region, it can be hard to tell whether shared features indicate a family or a sprachbund. The page says these shared features "might give a false appearance of relatedness."

AreaLanguagesShared features
BalkansAlbanian, Romanian, southern South Slavic languages, Greek, Balkan Turkish, Romani (Turkish is Turkic; the rest are Indo-European)avoiding the infinitive; forming the future tense in similar ways
South AsiaDravidian and Indo-Aryan, with Munda and Sino-Tibetan also cited (Emeneau, 1956)retroflex consonants, echo words, SOV word order, discourse markers, the quotative
Mainland Southeast AsiaSinitic, Hmong-Mien, Tai, Vietnamese, with Khmer, Cham, Laotone systems, similar vowel systems, isolating structure, classifiers, OV order, topic-comment structure; the modern consensus places these languages in unrelated families
Standard Average Europeanmodern Indo-European languages of Europe (Whorf, 1939); Romance and West Germanic at the core, North Germanic and Balto-Slavic more peripheral(the page says Proto-Indo-European lacked most of these features, so inheritance is ruled out)

The page also lists Mesoamerica as a linguistic area of the Americas, but gives no specific features. It attributes these similarities to diffusion during sustained contact and to convergence caused by language contact.

Note on sources
The wave-model, language-contact and linguistic-area material comes from Wikipedia's articles on the wave model, language contact and linguistic area. The wave-model page does not give an 1872 date for Schmidt or mention Schleicher, and the contact page does not mention Norse or give a borrowing hierarchy, so none of those appears here. I did not read these pages beyond the summaries returned. The Armenian-versus-Chapter-3 remark and the closing "tree for descent, waves for spread" are mine.

Try It Yourself

Exercise 1

In one sentence each, say what the tree model explains and what it does not, and what the wave model adds. Use the Germanic problem from the source.

๐Ÿ“„ View solution
Exercise 2

Greek, Romanian and Albanian avoid the infinitive, but so does Balkan Turkish, which is not Indo-European. Explain why the shared feature does not prove they are related, using the Balkans entry.

๐Ÿ“„ View solution
Exercise 3

Classify each as borrowing, pidgin or creole, mixed language, or language shift: Armenian taking Iranian words; Ndyuka and Saramaccan; Newar taking verb tenses from neighbours; English phrases such as attorney-general.

๐Ÿ“„ View solution

Chapter 6 Quick Reference

  • Tree = descent by splitting; wave = spread of an innovation; use both
  • Isoglosses map the boundary of a feature; linkages need the wave model
  • Contact outcomes: borrowing, code-switching, pidgin and creole, mixed language, shift and death
  • Heavy borrowing can fake relatedness (Armenian and Iranian); only basic correspondence sets prove kinship
  • A sprachbund shares features through contact, not descent (Balkans, South Asia, Southeast Asia)
Coming next
Historical Linguistics and Language Families 7 turns from sounds and families to meaning and grammar: semantic change, analogy and grammaticalisation.