Dissimilation¶
A phonological change or alternation in which one sound becomes less like a similar sound in a specified context.
Core Idea¶
Dissimilation is a sound-pattern change or alternation in which one sound becomes less similar to another relevant, initially similar sound within a specified phonological domain. A useful analysis identifies the pair, the shared feature, which sound changes, how that feature changes, and the word, root or morpheme environment in which it happens. The pair need not be adjacent, and neither “the first sound changes” nor “the weaker sound changes” is a universal rule. The outcome may be a changed segment or feature; the causal explanation is a separate research question.[1][2][3]
Latin and Greek/Sanskrit show why the relation deserves its own name without flattening their details. In Latin adjective formation, a stem /l/ can favor the rhotic suffix variant -āris instead of -ālis: mīles (stem mīlit-) gives mīlitāris. The later suffixal liquid changes category in this account. Grassmann's historical Greek/Sanskrit pattern instead concerns two aspirated consonants; one, generally the earlier aspirate, loses aspiration. The recurrent relation is context-linked reduction of similarity, not a fixed target direction or a fixed sound class.[1][2]
“Horror aequi” or avoidance of repetition is a possible motivation, not a defining operation. Müller reviews competing perceptual explanations for liquid dissimilation and documents Latin exceptions and distance effects; the observed conditioning pattern does not by itself prove that every speaker found repetition difficult or consciously avoided it. The label identifies the change relation while leaving mechanism and degree of regularity to evidence.[1]
Structural Signature¶
Sig role-phrases: similar sound pair in a stated domain → conditioned target → reduced shared feature in the output → language-specific scope and exceptions; repetition avoidance is a hypothesis, not a role requirement.
- Similar sound pair and domain. Two consonants or features share the property relevant to the case, such as lateral liquid quality or aspiration, within a specified word/root/morpheme domain. Without a similar trigger–target relation, an arbitrary sound substitution is not established as dissimilation. The relevant sounds can be separated by other segments.[1][2][3]
- Context-conditioned target. One member of the pair changes in the presence of the other under the pattern being claimed. Latin's suffixal liquid is later than the stem lateral; Grassmann's analysis generally removes aspiration from the first of two. Direction is case-specific, so a universal “earlier” or “less stressed” target is not constitutive.[1][2]
- Reduced shared feature. The target's output must be less like the similar companion in the relevant feature dimension: Latin suffixal /l/ versus /r/, or an aspirated versus unaspirated consonant. If a change instead increases similarity, it is assimilation; if similarity is unchanged, this defining outcome is missing.[1][2]
- Pattern scope and exceptions. Specify distance, intervening segments, morphology, language stage and attested exceptions before projecting a rule. Müller records an intervening /r/ condition and exceptions in Latin; Grassmann's historical law belongs to particular Greek/Sanskrit developments. These particulars make predictions honest but no one distance or exception pattern defines dissimilation universally.[1][2][3]
- Conditional causal story. Repetition avoidance, perceptual hypercorrection or another proposed pathway may explain a case. An observed context-linked divergence remains classifiable even if the historical cause is unsettled; treating a proposed motive as a compulsory structural role would preempt the evidence.[1]
What It Is Not¶
Dissimilation is not assimilation: assimilation makes sounds more alike in a context, while dissimilation reduces the relevant likeness. Nor does every /l/→/r/ or aspirate loss count. Without a comparable sound whose presence conditions the change, there may be a different sound change or a lexical variant but no demonstrated dissimilation relation. Merely finding -āris in one Latin word is weaker than demonstrating the -ālis baseline, conditioning stem lateral, and the pattern's exceptions.[1][3]
It is also not synonymous with the live Phonological Process entry, whose V2 identity specifically treats productive child speech errors, developmental suppression timing and clinical intervention. Historical Latin suffix allomorphy and Greek/Sanskrit aspirate change have neither that child-developmental target nor its diagnostic schedule. Dissimilation can be regular within a pattern or sporadic across individual words; it is not universally a disorder, a deliberate stylistic choice or a proven response to articulatory difficulty.[1][2]
Scope of Application¶
Latin liquid dissimilation has a particularly well-documented morphological setting. Müller identifies default adjectival -ālis, a stem-lateral-conditioned -āris pattern, and restoration of -ālis when an intervening /r/ occurs in the analyzed class. Her Table 1 also counts exceptions, so the abstraction cannot authorize “all /l/-containing stems require -āris.” Suzuki's typological dissertation gives sol-alis → solaris, lun-alis → lunaris, and milit-alis → militaris as illustrations, alongside blocking examples with intervening rhotics. The domain is the Latin adjective suffix and its documented environments, not every liquid anywhere in Latin.[1][3]
Grassmann's original 1863 discussion, available in a University of Texas-hosted English translation, concerns the distribution and historical treatment of aspirates in Greek and Sanskrit roots. Under the cited analysis, a root with two relevant aspirates loses aspiration in one, generally the first; the two language branches are treated as independent developments. This is a different feature and historical scope from Latin's lateral/rhotic suffix alternation. Other language cases require their own evidence rather than automatic extrapolation from either.[2]
Clarity¶
The contrast between description and explanation is essential. The descriptive claim asks whether a similar sound conditions a change that decreases similarity. The explanatory claim asks why the change occurred: speaker production burden, listener misperception, lexical reanalysis or another mechanism. Müller explicitly analyzes alternative accounts and warns through Latin exceptions that one surface distribution cannot settle every causal question. Naming the relation does not certify a psychological law.[1]
The target's position should likewise be read from the data, not from the term “dissimilation.” Latin suffixal -ālis → -āris affects the later liquid, whereas Grassmann's aspirate treatment generally affects the earlier of two. A compact rule should therefore state which feature diverges where, rather than assuming that dissimilation always deletes a segment, moves away from a neighboring sound, or targets the weaker member.[1][2]
Manages Complexity¶
A series of distinct word forms can look like unrelated lexical facts until their similar-sound context is compared. For the Latin adjective suffix, the common pattern links -ālis and -āris choices to the stem's liquid environment; it compresses many forms into a conditioned alternation. But because intervention and exceptions matter, the compression cannot be a rule that predicts every form from the existence of any /l/. The useful representation is a general condition plus a documented domain and limits.[1][3]
The cross-case concept compresses at another level: liquid rhoticization and aspirate deaspiration can be analyzed as unlike sound changes that reduce similarity between related units. That gives a common question to ask across languages without equating their sound inventories, morphological domains, directions or historical causes. The price of the common label is an obligation to preserve the feature and language-specific rule when making a concrete prediction.[1][2]
Abstract Reasoning¶
In the Latin suffix pattern, compare a baseline adjective-forming -ālis with a stem containing /l/. In the cited examples, the suffix liquid is realized as /r/, yielding -āris and reducing the two lateral occurrences to a lateral–rhotic pair. The mapping is not simply “/l/ always becomes /r/”: the affected liquid is in one derivational suffix, an intervening rhotic can alter the choice, and Müller's table records exceptions. The reasoning therefore moves from pair/context to changed feature and then checks the bounded lexical distribution.[1][3]
For Grassmann's historical aspirate pattern, the compared pair is not two laterals but two aspirated consonants. The original account describes loss of aspiration in one, generally the first, in the relevant Greek/Sanskrit historical configuration. The operation again reduces shared phonetic similarity, but its feature, position and domain differ from Latin. These two inferences justify the higher-order classification while refuting the seed's proposed universal earlier/unstressed target and repetition-avoidance cause.[2][1]
Knowledge Transfer¶
Transfer from Latin to Grassmann's law is literal only at the conditioned similarity-reduction relation: identify a similar pair, specify the target and feature, observe a less-similar output, and bound the pattern. The Latin outcome is suffixal liquid rhoticization in adjective formation; the Greek/Sanskrit outcome is aspirate deaspiration in a historical root setting. Carrying over Latin's later target or intervening-/r/ condition to Grassmann's case would be an error.[1][2]
The broader idea that similar elements may diverge in context might be examined as a future prime-like structural question. This entry does not show the same phonological identity across nonlinguistic substrates. It also cannot be imported into child-language pathology merely because live Phonological Process contains the word “phonological”; that entry has child target forms, developmental schedules and clinical use absent here.
Examples¶
Latin adjectival suffix. From mīles with stem mīlit-, the suffix yields mīlitāris rather than the default -ālis pattern in the cited analysis. Mapped back: pair/domain = a stem lateral and suffixal lateral within Latin adjective formation; target = later suffixal liquid; divergent feature = rhotic /r/ in -āris versus /l/ in -ālis; scope = intervening /r/ and documented lexical exceptions matter. The result is not a universal rule over every Latin word with /l/.[1][3]
Greek/Sanskrit aspirate pattern. Grassmann's original account reconstructs configurations with two aspirates and treats loss of aspiration in one, generally the first, as a historical development. Mapped back: pair/domain = two relevant aspirated consonants in the root/phonological configuration; target = generally earlier aspirate under this law; divergent feature = deaspiration; scope = Greek/Sanskrit historical systems and their conditioning, not Latin liquid allomorphy or all languages.[2]
Negative boundary. A sound becoming more like a neighboring sound is assimilation. An isolated suffix variant without demonstrated similar-sound conditioning may be lexical variation; the output alone does not establish dissimilation.[1]
Structural Tensions¶
Compact rule versus lexical fidelity. A stem-/l/ rule makes Latin suffix variation intelligible, but unqualified use misses intervening-/r/ behavior and exceptions. Listing every word separately preserves detail while hiding the recurrent condition. Diagnostic: Does this form belong to the relevant suffixal class, and what interveners or exceptional evidence affect it?[1][3]
Classified outcome versus proposed cause. The divergence can be described without knowing why it arose; a causal hypothesis may explain more but can also overclaim if rival perceptual or production pathways fit. Defining the phenomenon by “repetition avoidance” would exclude a supported surface relation if that motive failed, whereas refusing causal work would leave distribution unexplained. Diagnostic: Is the claim an observed alternation, or a separately tested account of its origin?[1]
Shared process versus feature-specific law. Grouping Latin liquids and Greek aspirates captures a genuine pattern of reduced similarity, but can obscure opposite target directions and distinct domains. Treating each as wholly unrelated preserves detail but forfeits the cross-case comparison. Diagnostic: Which pair, affected feature, target direction and language domain does the proposed rule actually cover?[1][2]
Structural–Framed Character¶
Evaluative weight. Dissimilation describes a sound relation, not an improvement in pronunciation or a desirable avoidance of repetition. Calling a form “less similar” does not judge it clearer or better; such evaluations require separate evidence.
Human-practice dependence. Linguists choose representations of /l/, /r/ and aspiration and define word/morpheme domains, while language users participate in the histories being described. Once a representation and evidence set are stated, whether a particular feature diverges in a conditioned environment can be checked independently of an analyst's preference.
Institutional origin. No language academy, classroom or clinic constitutes dissimilation. Latin morphology and Greek/Sanskrit historical aspirates fill the relation through different linguistic systems, and the phenomenon does not require a clinical developmental schedule.
Vocabulary travel. The term travels literally across phonological features and languages when a similar pair conditions a less-similar sound outcome. Metaphorical use for organizations “dissimilarizing” or broad social differentiation does not supply sound segments or phonological conditioning.
Import versus recognition. Recognition of a new case requires comparative forms, a trigger–target feature relation and a bounded domain. Importing an observed Latin /l/ rule into a new language without such evidence would substitute analogy for phonological analysis.
Its character: strongly structural within phonology, with a necessary language-specific feature/domain frame; proposed repetition-avoidance psychology is neither required nor established by the label.
Structural Core vs. Domain Accent¶
Portable skeleton. The abstract shape is context-conditioned divergence between initially similar elements. No checked live prime is verified as this whole skeleton: Contrast emphasizes existing difference rather than a conditioned sound change, and Transformation is broader and carries its own mapping/invariant requirements. A higher-order divergence concept remains a future-prime question, not an asserted parent.
Domain-bound mechanism. Speech sounds have analyzable features, positions, syllables and morphemes; Latin's suffix and Grassmann's historical root pattern depend on these carriers. Remove the sound pair, affected feature and language-specific domain, and one retains only a generic “becoming different” analogy, not phonological dissimilation.[1][2]
Why not prime. The two cited cases traverse unlike features and languages, but both remain phonological. Evidence does not show a single dissimilation identity in three unrelated nonlinguistic domains under the same trigger–target mechanism. The possible abstract divergence skeleton deserves separate adjudication rather than silently elevating this domain-specific entry.
Instantiates / Related Primes¶
This workspace proposes approved unparented placement. Live Phonological Process describes child-production transformations with age-graded suppression and clinical consequences; dissimilation here need not be a child error or have that developmental trajectory. Live Phonology is the discipline that studies sound systems, not a genus of this change. Contrast and Transformation are relevant broad comparisons, but the exact live signatures do not establish a necessary typed parent edge for every historical or alternational dissimilation. Placement may be revisited if a verified sound-change genus is admitted.
Neighborhood in Abstraction Space¶
Dissimilation sits in a sparse region of the domain-specific corpus (61st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Language, Mind & Meaning-Making (57 abstractions)
Nearest neighbors
- Folk Etymology — 0.86
- Rhyme — 0.85
- Perception of English /r/ and /l/ by Japanese speakers — 0.85
- Teaching prosody — 0.84
- Empty category principle — 0.84
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
Assimilation increases similarity; dissimilation decreases it under a specified relation. A random substitution has no demonstrated similar trigger. Horror aequi is one suggested explanation, not the operational identity. Grassmann's law is a narrower historical aspirate case, not the entire class. Phonological Process in the live encyclopedia is specifically child-developmental and clinical; the shared phonological vocabulary does not make that entry coextensive with historical sound change.[1][2]
References¶
[1] Daniela Müller, “Liquid dissimilation with a special regard to Latin” (2013), author-hosted manuscript, §2 examples (3)–(5) and Table 1, §§3 and 7 on competing explanations. https://www.phonetik.uni-muenchen.de/personen/assoziierte_wissenschaftler/mueller_daniela/publications/mueller2013liquiddissimilation.pdf registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y
[2] Hermann Grassmann, “Concerning the Aspirates and Their Simultaneous Presence in the Initial and Final of Roots,” 1863 original in Zeitschrift für vergleichende Sprachforschung 12(2), 81–138; English translation hosted by the University of Texas Linguistics Research Center in A Reader in Nineteenth Century Historical Indo-European Linguistics, Chapter 10. https://lrc.la.utexas.edu/books/reader/10-h-grassmann registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p
[3] Keiichiro Suzuki, A Typological Investigation of Dissimilation, University of Arizona PhD dissertation (1998), Rutgers Optimality Archive copy, PDF pp. 101–102, examples (46)–(47). https://roa.rutgers.edu/files/281-0998/roa-281-suzuki-1.pdf registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i