LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3708.
DOI: https://doi.org/10.56712/latam.v6i4.4543
Thunk: the cognitive processes behind language change
Thunk: los procesos cognitivos detrás de cambio en el lenguaje
.
Juan Carlos Rodriguez Burgos1
juarodrigez@uqroo.edu.mx
https://orcid.org/0009-0005-6366-0211
Universidad Autónoma del Estado de Quintana Roo
Cozumel – México
Artículo recibido: 06 de junio de 2025. Aceptado para publicación: 20 de septiembre de 2025.
Conflictos de Interés: Ninguno que declarar.
Abstract
This study investigates the cognitive processes underlying the emergence and usage of thunk as a
non-standard past participle and, less frequently, past tense form of think in present-day English. A
corpus-based qualitative approach was adopted, and data were drawn from the Corpus of Historical
American English (COHA), the Corpus of Contemporary American English (COCA), and the Oxford
English Dictionary (OED). A total of 71 relevant instances were identified, analyzed for diachronic
development, frequency, and contextual distribution. Findings reveal that while thunk historically
appeared in both past simple and participial contexts, its modern usage is restricted to the would have
+ participle and woulda + participle constructions, especially in colloquial expressions such as Who
would have thunk…?. As a result, the analysis considers three cognitive mechanisms (categorization,
analogy, and automatization) as potential explanations for this shift. Categorization links think to
phonologically similar strong verbs (e.g., drink – drunk), which enables analogical innovation. However,
evidence from corpus data suggests that automatization, driven by high-frequency co-occurrence and
phonetic reduction in predictable contexts, plays a central role in the entrenchment of thunk. These
findings highlight the importance of usage-based processes in language change and suggest that
thunk’s persistence is shaped more by routinized discourse patterns than by systematic morphological
reanalysis. Future research should explore phonetic realizations in spontaneous speech and
constructional frameworks to assess its integration into contemporary English.
Keywords: cognitive linguistics, analogy, automatization, corpus linguistics, language change
Resumen
El siguiente estudio analiza los procesos cognitivos que sustentan la aparición y uso de thunk como
una forma no estándar de pasado participio y, en menor frecuencia, de la conjugación en pasado del
verbo think en el idioma inglés contemporáneo. Se adoptó un enfoque cualitativo basado en corpus lo
que derivó en la extracción de información del corpus histórico del inglés americano (COHA), del
corpus del inglés americano contemporáneo (COCA), y del diccionario de inglés de Oxford (OED). Se
identificó un total de 71 ejemplos relevantes para la investigación, dando pie a su análisis de desarrollo
diacrónico, de frecuencia, y de distribución contextual. Dentro de los resultados se puede resaltar que
mientras que la forma thunk aparece históricamente en contextos donde se requiere su forma de
pasado participio y pasado simple, su uso contemporáneo, por el contrario, se ve confinado a
1 Autor de correspondencia.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3709.
construcciones del tipo would have + participio y woulda + participio, especialmente en expresiones
coloquiales como Who would have thunk…?. Consecuentemente, este análisis considera tres
mecanismos cognitivos que explicarían este cambio: la categorización, la analogía y la
automatización. La categorización vincula fonológicamente al verbo think con verbos de frecuencia
alta, drink – drunk, lo que permitiría una innovación analógica; sin embargo, los datos de los corpus
sugieren que la automatización juega un papel importante en el afianzamiento de la forma thunk,
debido a la coocurrencia de alta frecuencia y la reducción fonética en contextos predecibles. Estos
hallazgos subrayan la importancia de procesos basados en el uso dentro del cambio en el lenguaje y
sugiere que la persistencia de la forma thunk está moldeada por patrones de discurso rutinizado en
lugar de un reanálisis morfológico sistemático. Se espera que futuras investigaciones exploren las
ejecuciones fonéticas en el habla espontánea y marcos construccionales para evaluar su
incorporación en el inglés contemporáneo.
Palabras clave: lingüística cognitiva, analogía, automatización, lingüística de corpus, cambio en
el lenguaje
Todo el contenido de LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades,
publicado en este sitio está disponibles bajo Licencia Creative Commons.
Cómo citar: Rodriguez Burgos, J. C. (2025). Thunk: the cognitive processes behind language change.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades 6 (4), 3708 – 3721.
https://doi.org/10.56712/latam.v6i4.4543
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3710.
INTRODUCTION
Languages evolve in unique and often unexpected ways: irregularities can gradually enter a language,
with users often unaware that previously standard grammatical or lexical forms are being replaced. But
how do these irregularities find their way into the language?
Two major perspectives can be identified in addressing this question: the dictionary-and-grammar view
of language (Taylor, 2012) and the constructionist perspective of linguistic knowledge (Goldberg,
2003). The former is concerned with a clear-cut distinction between grammar, a set of specific rules on
any given language, and a second element in charge of comprising all irregularities and vocabulary. In
contrast, the latter proposes that no distinction is made, rather knowledge of a language emerges
through cognitive processes involved in language use. Consequently, generalizations of constructions
are recognized subconsciously by language users and these are stored in what is known as the
construct-i-con (Hilpert, 2014)
This paper aims to investigate the cognitive processes underlying the use of thunk, a non-standard past
tense and participle form of think, in present-day English. The phenomenon is examined as an instance
of language change driven by cognitive mechanisms.
The paper first contextualizes the case within existing literature on frequency effects and cognitive
linguistics. It, then, explores the historical emergence of thunk, and examines dictionary attestations,
such as the Oxford English Dictionary (OED), and corpus data to determine its emergence, frequency
and contextual usage. Finally, it discusses the cognitive processes that may explain its development,
including categorization, analogy, and automatization. Conclusions are drawn regarding the nature of
the change and potential areas for future research.
METHODOLOGY
This study employs a corpus-based qualitative analysis to trace usage patterns of the non-standard
verb form thunk in present-day and historical English. As a result, two corpora were consulted: the
Corpus of Historical American English (COHA) and the Corpus of Contemporary American English
(COCA). Each occurrence was manually examined, so a total of 58 relevant instances were extracted
from COCA and 13 from COHA. Additionally, the Oxford English Dictionary (OED) was consulted to
check whether it had already been established in the language as well as to trace the diachronic
development and lexical status. Finally, findings were interpreted through the lens of cognitive linguistic
theory, particularly in relation to categorization, analogy, and automatization. No statistical tests were
applied, as the analysis was exploratory and interpretive in nature.
DEVEPOLMENT
Previous studies
Frequency effects
“(…) Frequency of occurrence is an important determinant of linguistic structure and language use. (…)
Frequency has an impact on the comprehension, production, and emergence of linguistic categories
and rules (Diessel, 2007, p.109)”. For this reason, corpus-based analyses are necessary to determine
whether a linguistic change is influenced by frequency.
Gahl and Garnsey (2004) investigated the frequency and probabilistic effects of /t, d/ deletion in
environments that were biased towards specific syntactic structures. The results were categorical: the
pronunciation of regular past tense verbs (e.g., confirmed, suggested, etc.) was affected by the
syntactic context. Verbs that typically occurred with a direct object (DO) or a sentential complement
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3711.
(SC), as shown by corpus data, exhibited greater phonetic reduction in the past tense when followed by
a DO rather than an SC. Thus, it can be assumed that frequency of collocational patterns influences
phonetic realization.
Hooper (1976) also highlighted how phonological reduction is affected by lexical frequency. In her
research, schwa deletion occurred more frequently in common lexical items (e.g., memory, salary,
summary and nursery), especially when preceded by a sonorant sound (preferably /r/ or /l/), compared
to less frequent items (e.g., mammary, artillery, summery and cursory). Frequency counts provided by
Francis and Kucera (1982) confirmed this pattern: for example, every appeared 492 times, while artillery
only 11.
Bybee (1985, 1988) further argued that frequency of use influences the way words are stored and
accessed. High-frequency items become deeply entrenched in memory, which leads to faster retrieval.
Moreover, irregular word forms are maintained in the language as long as they exhibit high token
frequency; for example, was/were. In contrast, irregular forms that occur infrequently undergo
regularization over time; for example, burnt �� burned.
Cognitive processes
Bybee (1995) compares three models of morphological representation: the dual-processing model, the
Connectionist model and the Network model. Morphology is, thus, conceptualized as an emergent
property of associative networks based on lexical, morphological and semantic connections, where
frequency also plays a central role. Bybee notes that “sets of words having similar patterns of semantic
and phonological connections reinforce one another and create emergent generalisations describable
as schemas” (1995, p. 430). Therefore, when a new or infrequent item is introduced or retrieved, it will
be compared and fitted into one of these schemas as long as it aligns with the schema’s properties
and strength (i.e., the more items a schema possesses, the stronger it is).
De Smet and Fischer (2016) highlight the importance of analogy in language change as it is linked not
only to categorization but also to broader domains such as cognitive science and language acquisition.
They highlight the role that frequency plays in analogical extension and how it is connected to
categorization, which other authors have already discussed (Bybee & Beckner, 2015). Their analysis of
the grammaticalization of have to and as good as illustrates how supporting constructions influence
analogical spread.
Bybee (2003) adds automatization as a crucial mechanism in grammaticalization. According to usage-
based theory, frequent repetition of structures such as going to lead to phonological reduction and
chunking, ultimately resulting in lexicalized units (e.g., gonna). Automatization reduces motor effort and
leads to opacity; that is, the original components lose their transparecy and are no longer perceived as
compositional.
Thesis
Thunk is most frequently used as a past participle and is primarily restricted to the context of would
have + participle. Its development is likely the result of cognitive processes, categorization and analogy,
which are involved in language use.
RESULTS
Emergence of thunk
To determine whether thunk has been formally established in English, the Oxford English Dictionary
(OED) was consulted. It lists four entries for thunk, of which only one is relevant to this study:
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3712.
“Joc. variant of ‘think n.’ (OED: thunk, n.1)
“A sound of an impact, either dull or plangent. Also int. or as adv. (OED: thunk, n2)
“Dial. and joc. pa. tense and pa. ppl. of ‘think v.’”(OED: thunk, v1.)
“intr. To make a thunk; to fall or land with a thunk” (OED: thunk, v2)
The first, second and fourth definitions were excluded as they do not pertain in the verbal alternation
under analysis. The third entry, which concerns the verbal form, is marked with frequency band 3 in the
OED. According to the dictionary’s frequency guide, this represents relatively rarity, but not
obsolescence.
The earliest attested use, dated to 1876, describes thunk as being more common in the past tense than
in the participle form:
Think. (Thuongk). The last form is less employed participially than in the past, in which tense it is of
constant occurrence.” (C.C. Robinson, 1876, as cited in Davies, 2010)
This observation challenges the central hypothesis of the present study, which posits a higher
frequency of thunk in participial contexts. Unfortunately, the limited number of examples in the OED
(only four related to this usage: two in the past tense and two in the participle) is insufficient to confirm
or reject the hypothesis conclusively.
If the earliest attested form is considered, a different historical trajectory must be acknowledged, one
that suggests thunk underwent semantic or functional shifts over time.
COHA
The Corpus of Historical American English (COHA) provides 135 occurrences of thunk between 1810
and 2000. However, not all instances reflect the verbal form under investigation. A detailed examination
was conducted to identify only those cases where thunk functioned as either a past tense or past
participle of think:
“(…) I'd be a butterfly born in a bower, Where roses and -- " Fitz. Fitz James Why, who'd have thunk it --
so good and yet so bad (…)” (The Lady of the Lake, 1861, as cited in Davies, 2010)
“(…) but she leans up to him purty sharp, I reckons. Dar was one time Ise thunk Marsa Hampton had'
em all fixed; but sence he's done gone to de Debil of de Forest, Marsa Colbrum' pears to hev it all cl'ar
wedder (…)” (The Phantom of the Forest, 1868, as cited in Davies, 2010)
These early instances confirm that thunk historically appeared in both grammatical roles. Notably, the
participial form was not only restricted to the canonical structure would have + participle, but it was
also used in past perfect constructions but to a lesser degree:
“(…) I knew when my Reginald staggered into the door-yard that he was on the Die, but if I'd only
thunk to ask him about them mules ere his gentle spirit took flight, it would have been four thousand
dollars in our pockets, and no mistake! (…)” (Artemus Ward; His Travels, 1865, as cited in Davies, 2010)
Over time, such variation diminished, and thunk has become almost exclusively associated with the
would have + participle structure in contemporary usage. This restriction is further supported by the
latest examples in COHA, which reflect a narrowing of contexts as opposed to early instances that
could be either as a past form or a participle.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3713.
There is a time gap when it comes to the penultimate and last usage of thunk in the past simple form.
However, this might just reflect the potential lack of information that could help track its history:
“(…) ‘He wouldn’t dare to sell them out.’ Well, I thunk some more thinks just as punky as that, and then
we settles it that I’m to hike over and take a squint, anyway. (…)” (Torchy, 1911, as cited in Davies, 2010)
“(…) After you gone to Cambry this morning I thunk on it and I know it I bes have another look at things
in Fork Stoan (…)” (Riddley Walker, 1980, as cited in Davies, 2010)
By analyzing the occurrences longitudinally, a noteworthy trend arises: COHA data shows a peak in
participial usage during the 1880s (10 instances), followed by a marked decline. From 1890 to 1970,
the corpus registers only a handful of relevant instances, often just one per decade, with none in the
1940s. The 1980s and 1990s show a modest resurgence, with three and two occurrences respectively,
and only one case is found in the 2000s. Moreover, COHA started recording thunk as a noun shortly
after example number (4). This might have contributed to its less frequent usage in the past simple
form as this lexical overlap may have led to some confusion or underrepresentation of the verbal form:
“(…) The howling on the banks of the little stream was now almost deafening, and every second there
came the thunk of arrows against the boat (…)” (Boy Scouts in Airship, 1912, as cited in Davies, 2010)
In summary, COHA provides useful diachronic evidence which supports the idea that thunk was indeed
present in both past tense and past participle in the history of the language, but it fails to capture its
modern-day usage. To provide more current data, it is necessary to consult a contemporary corpus,
which is addressed in the following section.
Frequency of usage: COCA
A final corpus study was conducted to corroborate the usage of thunk in the OED. As previously
mentioned, the OED classifies its frequency as belonging to band 3, which indicates relatively rarity.
The results showed that its frequency is indeed quite low, with only 289 occurrences found. Upon closer
examination, it became evident that these instances included not only the verbal form in question but
also nominal and onomatopoeic uses. Therefore, an exhaustive analysis was carried out to isolate only
the relevant tokens. A total of 58 examples were identified as legitimate verbal instances of thunk,
which confirms that it is a highly infrequent alternation of thought.
Despite the low frequency, COCA helped determining the contexts in which thunk is used. Interestingly,
it contradicts the OED’s claim that thunk functions as both past tense and past participle. It is worth
noting that, in COCA, thunk appears to be predominant in the structure would have + participle, and
more colloquially, woulda + participle.
Context of thunk
The OED describes thunk as both a past simple and a participial form; however, it provides insufficient
evidence to determine whether it shows preference for one over the other. Nonetheless, the definition
serves as a useful foundation for further exploration in trusted corpora. Notably, the participial form
appears to be restricted to the context of would have + participle and, even more specifically, to the
reduced form woulda + participle, as only such instances are shown in the online dictionary. This
challenges the original thesis of this paper, as it suggests that the form may be the result of a different
cognitive process not previously considered: automatization (see section 7.3).
Historical corpus data offers a broader view of thunk’s contextual usage. The variant appears in five
main constructions: 1) would have + participle, 2) past perfect, 3) woulda + participle, 4) past simple,
and 5) musta + participle.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3714.
COHA suggests that thunk preferred the past simple context, with 13 relevant instances in the corpus.
However, when all examples which involve the participial form are taken into account, thunk shows a
strong preference for the participle, particularly in combination with would have. Moreover, when
comparing the would have + participle and woulda + participle variants, a surprising pattern arises:
thunk appears more often in the reduced form context (woulda, 10 instances) than in the full form.
Interestingly, would is the preferred modal, with only one example that involves a different modal verb
found in 1981:
(…) So proud, she must a thunk she was in Longchamps (…) (Grown Ups, 1981, as cited in Davies, 2010)
This suggests a strong tendency for thunk to occur in contexts in which have becomes reduced and it
is preceded by a modal verb. If this pattern is observed in a contemporary corpus data, its evolution
pattern can be assessed.
COCA displays different trends in the usage of thunk. First, there were no instances involving modal
verbs other than would, unlike COHA. This implies that while thunk briefly extended to new modal
contexts in the past, these innovations were not sustained. Had it been successful, we would expect a
higher number of instances with alternative modals such as must, could, or others.
In contrasts to COHA, COCA shows five occurrences of thunk in simple past, which contradicts previous
findings. This discrepancy might be due to iconicity, since the same form functions as a noun
representing a sound. As Wang (2010, p.177) explains:
“An icon image is a single sign that resembles its referent with respect to its characteristics (…). In a
word, iconicity as a semiotic notion refers to a natural resemblance or analogy between the form of a
sign (…) and the object or concept (…) it refers to in the world or rather in our perception of the world.”
In other words, once thunk emerged as a noun describing a sound, the language may have avoided
using it in the simple past to reduce ambiguity, thus favoring its use in the participial context. The
auxiliary have reinforces the link with the base verb think, which helps maintain clarity. The overlap
between the nominal and verbal forms may have contributed to the decline of the past simple usage.
Particularly, COCA does not show thunk in other tenses, unlike COHA. As stated previously, languages
evolve in peculiar ways, and those changes are motivated by cognitive processes that cannot be
controlled.
There is now little doubt that thunk most frequently appears in the would have + participle structure.
However, the question remains: is it more frequent in full would have + participle form or in the reduced
woulda variant? The corpus data yield contradictory results. While COHA shows more examples of the
reduced form (10 vs. 2), COCA displays the opposite: 32 instances of would have + thunk and 18 of
woulda + thunk. This makes it difficult to draw firm conclusions. One explanation is that the reduced
form was more common in earlier periods, while the full form gained ground over time. Another
possibility, and perhaps the more credible, is that this discrepancy reflects differences in genre and
time coverage between the corpora.
Despite this, the overall pattern is clear: thunk is typically found alongside would have, especially Wh-
question such as Who would have thunk…?. As previously proposed, thunk likely emerged through
cognitive processes that are involved in everyday language use. Nonetheless, the analysis has revealed
an additional pathway of change that was not considered at the beginning. One sub-thesis, though, has
been supported: thunk is more frequent in participial contexts than in the past simple.
Cognitive processes involved in the development of thunk
Categorization
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3715.
Structuralist and generativist views of linguistic knowledge differ from the notion of the construct-i-con
in that “material contained in rules does not also appear in the lexicon and vice versa” (Bybee, 2001, p.
20). Theforefore, these models assume that a dedicated mental module handles words and
idiosyncracies, while another independent module manages grammar. In contrast, cognitive linguistic
approaches maintain that there is a close and dynamic relationship between lexicon and grammar,
which makes the latter view more plausible and cognitively economical.
Langacker (1987) discusses the example of the -ed suffix: a child must be exposed to a number of
instances (e.g., spilled, played, talked) to form generalizations over concrete material. In other words,
“our brains record even the predictable pieces of the experience” (Bybee, 2001, p. 20). This reflects a
subconscious capacity that enables speakers to recognize similarities across words and phrases.
Language users can store new material based on recognized features that were derived from previous
language use, whereas completely new features may create new memories or generalizations.
Bybee (2001, p. 21) provides an analogy: “if our memories for dogs exluded all the predictable features
(two ears, a muzzle, fur, a tail, wet nose, etc.) what is left would not be a recognizable or coherent entity.”
Consequently, speakers’ knowledge of a word or phrase includes all those predictable features that can
be cognitively retrieved. When language users are exposed to the verb think, for example, they can
associate it with other words that share phonological features (e.g., wink, link; three, theme) or semantic
associations (e.g., believe or consider). Additionaly, the aim is not to dismiss the concept of regularity,
but to acknowledge that categorization allows similarities to surface naturally.
If thunk is analyzed at a morphological level, language users should be capable of recognizing
phonological similarities in think that are comparable with verbs such as sing, drink and ring. Hence, a
reinterpretation of its past forms could be prompted, given the partially similar patterns. These
similarities can be visualized as follows:
Figure 1
The rise of morphology: identifying morphological similarities between words
The morphological form of think may be cognitively categorized with other verbs based on phonological
resemblance. Some relationships show a strong resemblance (think – drink), while others are weaker
(think – ring, think – sing). As a result, these patterns suggest that such items are stored in proximity
in the mental lexicon. Despite not having a clear-cut morpheme that could link these words, shared
phonological traits can support the mental organization of irregular morphology.
Figure 1 conceptualizes what Bybee (1985) considers to be an improvement over the idea presented in
the Phonetic refinement theory (Pisoni, Nusbaum, Luce, & Slowiaczek, 1985) as “ (…) identity relations
are represented by connecting lines. (…) These lexical connections can be phonological or semantic
(…)” (Bybee, 2001, p. 22). As previously noted, these connections resemble what other morphological
theories would describe as prototypes. Hence, there is no need for an independent module to manage
irregular morphology; categorized material can be stored as schemas, even if these representations
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3716.
lead to redundancy, as words that can be easily described by predictable morphemes could also be
stored independently.
But why is this process of categorization important? Not only does it provide a cognitively plausible
account of morphological structure, but it also offers insight into the dynamics of change. Words that
are stored closely together in the mental lexicon may influence one another, which could potentially
lead to reanalysis or analogical innovation.
Analogy
McColl Millar & Trask (2015, p. 99) provide a useful example to illustrate the logic behind analogy:
imagine being tested on the highly infrequent nouns ziff, zo and zax to write their plural forms. The
answer would likely be ziffs, zos and zaxes, respectively. Speakers, thus, are able to produce such forms
through anology; that is, speakers apply familiar patterns, in this case, the regular plural suffix -s, to new
or unfamiliar items.
Analogy, therefore, is the process in which speakers can create new forms by drawing parallels with a
number of existing forms, regardless of being a large or small quantity. As a result, anology is
considered “a powerful mechanism for language change” (McColl et al, 2015, p. 99), especially in
morphological developments. Put it in simple terms, a speaker is likely to be familiar with the language
pattern A-B, and once they are exposed to a new or an infrequent word in everyday speech, such pattern
is readily available because of its high degree of productivity or because the word possesses traits that
resemble another.
There are two primary types of analogy: extension and levelling. The former involves applying a known
pattern to new or rare items. McColl Millar and Trask (2015) provide an example regarding loanwords:
there is a number of nouns in English that were borrowed from Latin, whose plural form follows the -i
suffix source pattern (e.g., cactus-cacti, radius-radii, succubus-succubi). When octopus, a Greek
borrowing, entered English, speakers noticed the -us ending and erroneously extended the Latin plural
pattern, resulting in octopi, despite the correct Greek plural, octopode.
Analogical levelling, on the other hand, describes a situation in which two or multiple morphological
variants of a word coexist, but eventually one form predominates and replaces the others. Haspelmath
(2002) illustrates this process through the Old French verb trouver and its two stems alternations. When
the verb was stressed on the final syllable (as it is with the pronouns nous and vous), the stem took the
form ou. When the stem itself was stressed (as with pronouns je, tu, il, elle, ills, elles), it appeared as eu.
This alternation no longer exists in modern French, so ou became the default. Overall, analogical
levelling eliminates stem alternations allowing the language to show greater consistency.
Both processes can be illustrated in the form of equations. In the case of thunk, this analogy might look
like the following:
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3717.
Figure 2
Proportional equation representing the formation of thunk
As shown, thunk could have emerged by analogy with strong verbs like drink - drunk. However, it would
be misleading to attribute the change solely on one trigger word such as drink. Instead, the change
likely stems from shared patterns, word-schemas, across semantically and phonologically similar verbs
(Haspelmath, 2002).
Addtionally, think does not conform to the full strong-verb- pattern A-B-C (e.g., sing – sang – sung or
drink – drank – drunk). There is no evidence for a hypothetical past tense form *thank in the corpora
examined, only think-thunk-thunk had been registered in the OED as previously stated. This likely
reflects a cognitive preference for clarity and distinctiveness, as languages prefer to be as transparent
as possible; a concept known as iconicity (Haspelmath, 2008). If there were, indeed, examples where
*thank was used as the simple past form of think, this would create confusion with words already in
the lexicon but with complete different meanings (e.g., thank; meaning of gratitude). This is a peculiarty
that languages tend to avoid.
McColl Millar and Trask (2015, pp. 100-101) give a phonological example: English lost the /w/ sound
after /s/ and before /o/; as a result, sword would receive a silent /w/ sound, but the spelling remained
unaffected. The same change was expected with swore and swollen, but it retained the /w/ sound. This
may have happened either bacause their base forms (swear and swell) influenced their pronunciation
to preserve morphological transparency, or because the /w/ sound was restored analogically. In any
case, the phenomenon reinforces the idea that analogy operates not in isolation but within a network
of related patterns.
Automatization
At a subconscious level, speakers tend to respond effortlessly to words or structures that occur
frequently in their linguistic environment. As listeners, certain pronunciations might be perceived as
‘lazy’ or ‘sloppy’ when speakers reduce or simplify articulation. These patterns are not sign of slackness,
but the result of a cognitive process that allows speakers to communicate more efficiently with reduced
articulatory effort; a concept known as automatization, also referred to as routinization.
As Hilpert (2014, p. 144) explains, “you are more likely to articulate your words carefully in situations
that are formal (…) or when these is ambient noise that impairs auditory comprehension.” In other
words, speakers tend to adjust their speech to maintain clarity and meet social expectations in formal
settings. However, there are other factors, linguistic and contextual, which contribute to speech
reduction.
Crucially, frequency and predictability are key linguistic variables that influence this process (Hilpert,
2014, p. 144). He illustrates this with the word and, a highly frequent conjunction that, as a result of its
frequency, undergoes phonetic reduction. Moreover, given the predictability of certain collocations,
listeners can infer and in its reduced form. For example, when exposed to the phrase ‘gin ____ tonic’,
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3718.
most people can effortlessly predict the missing word. Therefore, frequency and predictability are both
important factors in phonetic reduction.
Why was it necessary to define the context of thunk at the beginning of this study? As shown in the
examples from the OED, the original hypothesis fell short: specifically, a case of automatization
emerged upon analyzing the recorded instances. As it had been clarified in previous sections, thunk is
most commonly present in the structure would have + participle, replacing the standard form thought.
Additionally, corpus data indicate that thunk frequently appears in the structure who would have
thunk…?
DISCUSSION
Pathways of change
On the one hand, the emergence of thunk can be the result of both categorization and analogy, as
initially proposed. The way speakers access and use language in everyday situations allows think to be
associated with characteristics shared by certain strong verbs. This cognitive alignment may have led
think to shift from its standard past tense and participle form (thought) to resemble the pattern of verbs
such as sing, drink, and ring.
Nevertheless, one potential counterargument is that its past tense did not change independently, as the
hypothetical pattern think-*thank-thunk never materialized. However, this does not invalidate the
analogical process. Even though the pattern A-B-C (e.g., sing – sang – sung) was not the driving force,
it is more likely that the base and participial form of such schema may have influenced the change.
Consequently, the language adopted the pattern think – thunk – thunk as an analogical variation.
Another possibility is that the original A-B-B pattern (think – thought – thought) was so deeply
entrenched that the simple past resisted change, or the analogical pressure was insufficient to alter
both forms.
On the other hand, automatization might offer another explanation. In the historical form thuongk
recorded in the OED (see section 5), thuongk had a more frequent use in the past form than as a
participial. Given this increased usage, the time required to produce the form decreased, which may
have led to its phonological reduction to eventually become routinized. Simultaneously, the reduction
of would have to woulda could have facilitated the chunking of woulda + thunk, especially since these
elements frequently co-occur in discourse (Who woulda thunk…?) as previously discussed.
This co-occurrence of would, have and thunk in highly predictable contexts has led to the
automatization of the full sentence. The fact that this process can be grounded in frequency and
cognitive processes, it is assumed that the emergence of thunk is through usage-based mechanisms
rather than grammatical reanalysis.
Although the original thesis emphasized categorization and analogy, the analysis ultimately suggests
that automatization plays a central role in the development of thunk. Additionally, further research
should focus on phonetics and sociolinguistics of contemporary spoken data to determine how thunk
behaves in real-time language use.
CONCLUSION
Kiparsky (1995) refers to language change as often being the result of analogical processes as
linguistic forms shift from one pattern to another, eventually spreading to a broader set of items. In this
view, language change is inherently connected to the principles of cognitive linguistics.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3719.
This paper has argued that changes in language, specifically the emergence of thunk, can be accounted
for through cognitive processes such as categorization, analogy, schema formation, and
automatization. The thesis, initially, focused on the influence of categorization and analogy in reshaping
the morphological pattern of think. However, as the analysis progressed, particularly through corpus
evidence and dictionary attestations, it became clear that automatization may offer a more compelling
explanation for the evolution of thunk.
As speakers use language in daily interactions, we start recognizing patterns that help us store
linguistic material that is either new or already in existence. The form thunk may have originated
through analogy with other strong verbs, but its integration into everyday speech appears to have been
strongly influenced by chunking, especially in constructions such as who would have thought…?,
resulting in the form who woulda thunk?. This suggests that high-frequency co-occurrence and phonetic
reduction played key roles in the form’s entrenchment.
An area for further research would be to apply Hilpert’s (2014, pp. 14-22) framework for identifying
constructions, to determine whether thunk qualifies as part of a larger constructional schema.
Additionally, it would be worth analyzing speaker pronunciation in spontaneous speech, particularly of
the phrase Who would have thunk…?, as it could reveal how the segment would have contributes to the
routinization and phonetic shaping of thunk in discourse.
Regardless of the precise pathway, whether analogy, automatization, or a combination of both, this
paper has sought to demonstrate that thunk can be plausibly explained through the lens of cognitive
processes involved in language use. In this respect, language change should be examined not only
synchronically, in terms of current use, but also diachronically, tracing the historical and cognitive
developments that shape linguistic evolution.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3720.
REFERENCES
Bybee, J. (1985). Morphology: A study of the relation between meaning and form. John Benjamins.
Bybee, J. (1988). Morphology as lexical organization. In M. Hammond & M. Noonan (Eds.), Theoretical
morphology: approaches in modern linguistics (pp. 119-141). Academic Press.
Bybee, J. (1995). Regular morphology and the lexicon. Language and cognitive processes, 10(5), 425-
455.
Bybee, J. (2001). Phonology and Language Use. Cambridge University Press.
Bybee, J. (2003). Cognitive processes in gramaticalization. In M. Tomasello (Ed.), The New Psychology
of Language (Vol. 2, pp. 145-167). Lawrence Earlbaum Associates.
Bybee, J. (2012). Domain-general processes as the basis for grammar. In M. Tallerman & K. R. Gibson
(Eds.), The Oxford Handbook of Language Evolution (pp. 528-536). Oxford University Press.
Bybee, J., & Beckner, C. (2015). Language use, cognitive processes and linguistic change. In C. Bowern
& B. Evans (Eds.), The Routledge Handbook of Historical Linguistics (pp. 503-518). Routledge.
Davies, M. (2008-). The Corpus of Contemporary American English (COCA): 520 million
words, 1990-present. Retrieved from http://corpus.byu.edu/coca/
Davies, M. (2010-) The Corpus of Historical American English (COHA): 400 million
words, 1810-2009. Retrieved from http://corpus.byu.edu/coha/
De Smet, H., & Fischer, O. (2016). The role of analogy in language change: supporting constructions. In
M. Hundt, S. Mollin, & S. Pfenninger (Eds.), The changing English language: Psycolinguistic perspectives
(pp. 240-268). Cambrige University Press.
Diessel, H. (2007). Frequency effects in language acquisition, language use, and diachronic change.
New ideas in psychology, 25(2), 108-127. http://www.personal.uni-
jena.de/~x4diho/Frequency%20effects.pdf
Francis, W. N., & Kucera, H. (1982). Frequency analysis of English usage: Lexicon and grammar.
Houghton Mifflin.
Gahl, S., and Garnsey, S. M. (2004). Knowledge of grammar, knowledge of usage: Syntactic
probabilities affect pronunciation variation. Language, 80(4), 748-775.
Goldberg, A. E. (2003). Constructions: A new theoretical approach to language. Trends in Cognitive
Sciences, 7(5), 219-224.
Haspelmath, M. (2002). Understanding Morphology. Arnold.
Haspelmath, M. (2008). Frequency vs. Iconicity in explaining grammatical asymmetries. Cognitive
Linguistics, 19(1), 1-33.
Hilpert, M. (2014). Construction Grammar and its Application to English. Edinburgh University Press.
Hooper, J. B. (1976). Word frequency in lexical diffusion and the source of morphophonological change.
In W. Christie (Ed.), Current progress in historical linguistics, (pp. 96-105). North-Holland.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2025, Volumen VI, Número 4 p 3721.
Kiparsky, P. (1995). Explanation in Phonology. In J. A. Goldsmith (Ed.), The Handbook of Phonological
Theory (pp. 640-670). Blackwell.
Langacker, R. W. (1987). Foundations of Cognitive Grammar: Theoretical Prerequisites (Vol. 1).
Stanford University Press.
Lindblom, B., MacNeilage, P., & Studdert-Kennedy, M. (1984). Self-organizing processes and the
explanation of language universals. In G. Butterworth, B. Comrie, & Ö. Dahl (Eds.), Explanations for
language universals (pp. 181-203). Walter de Gruyter.
McColl Millar, R., & Trask, L. (2015). Trask's Historical Linguistics (3rd ed.). Routledge.
Oxford University Press. (n.d.). Key to frequency: calculating frequency. Oxford English
Dictionary Online. http://public.oed.com/how-to-use-the-oed/key-to- frequency/
Oxford University Press. (n.d.) Thunk. Oxford English Dictionary Online.
http://www.oed.com/search?searchType=dictionary&q=thunk&_searchBtn =Search
Pisoni, D. B., Nusbaum, H. C., Luce, P. A., & Slowiaczek, L. M. (1985). Speech Perception, Word
Recognition and the Structure of the Lexicon. Speech Communication, 4(1-3), 75-95.
https://doi.org/10.1016/0167-6393(85)90037-8
Saussure, F. (1986). Course in General Linguistics (C. Bally & A. Sechehaye, Eds.; R. Harris, Trans.).
Open Court. (Original work published 1916).
Taylor, J. R. (2012). The Mental Corpus: How Language is Represented in the Mind. Oxford University
Press.
Wang, X. (2010). Language Iconicity from Sociolinguistic Perspective. Asian Social Science, 6(7), 176-
179.
Todo el contenido de LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, publicados en
este sitio está disponibles bajo Licencia Creative Commons .