12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 1 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

Subject

Key-Topics

DOI:

25. Psycholinguistic Perspectives on Language Change

JEAN AITCHISON

Linguistics

»

Historical Linguistics

language

10.1111/b.9781405127479.2004.00027.x

Historical linguists aim to explain language change, ultimately in terms of properties of the human mind. A
massive amount of work on causes of change, therefore, could be regarded as “psycholinguistic,” or in
broader, more fashionable terminology, “cognitive.” In practice, a limited number of topics have been
highlighted as belonging to this domain, even though dividing lines between psycholinguistic/cognitive and
other fields of enquiry are fairly difficult to draw: “In no way can a pragmatic account be usefully separated
from a cognitive one” (Payne 1992: 3) is a typical comment.

Psycholinguistic/cognitive discussions of change are essentially about “top layer” causation. In historical
linguistics, at least three overlapping layers of causes can usefully be distinguished. First is immediate
trigger, which can be regarded as “sociolinguistic,” as when, in Labov's seminal paper, the permanent
inhabitants of Martha's Vineyard imitated the vowels of the fishermen they subconsciously admired (Labov
1963). Second is “linguistic proper,” due to vocal tract configurations, or to the maintenance of language
patterns, as when, in the Martha's Vineyard case, the diphthongs [ai] and [au] follow a roughly parallel course
of change. The third level is when broad properties of the human mind are hypothesized to account for any
such changes. For example, memory limitations or processing procedures might plausibly be invoked to
explain why front and back vowels tend to move in tandem not just in Martha's Vineyard, but everywhere.
This “top layer” of causation can be labeled psycholinguistic or cognitive, even though such labels need to
be used with care: they have been applied in the literature in a number of different ways. In this chapter,
psycholinguistic and cognitive are used in the fairly broad sense of “relating to language and mind in a way
which underlies or goes beyond strictly linguistic explanations.” This territory is a no-person's-land between
language universals and more general psychological ones. The location of the boundary is often dependent
on the theory adopted, rather than on an (unachievable) Olympian view of the situation.

Within this fuzzy language and mind area, two broad topics have been repeatedly linked to questions of
language change: one is child language acquisition, the other is speech processing. This chapter will outline
some of the basic controversies, and assess the state of play.

1 Child Language

Each child has to create language afresh for itself. This mundane truism has given rise to a recurring belief
that change occurs between generations. It was popular at the end of the nineteenth century: “The chief
cause of sound changes lies in the transmission of sounds to new individuals,” stated Hermann Paul (1880:
63), for example. This view still recurred in the twentieth century: “The ultimate source of … linguistic
change … is the process of language acquisition” (Andersen 1978: 21); “A basic cause of change is the way
children acquire the language … The child's grammar is never exactly like that of the adult community”
(Fromkin and Rodman 1993: 348).

12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 2 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

The belief in adult-child language discontinuity is therefore a longstanding one. However, the mechanism
which is presumed to underlie this generation gap varies with the decades. Imperfect learning by infants was
a favored mechanism in the nineteenth century: “If languages were learnt perfectly by the children of each
generation, then languages would not change,” asserted Henry Sweet (Sweet, 1899: 74).

In the 1960s, children's natural language learning ability was opposed to adult limitations. The years 2–14
were held to constitute a “critical period” for language acquisition, and its “termination was related to a loss
of adaptability and inability for reorganization in the brain” (Lenneberg 1967: 179). Halle (1962) was
possibly the first explicitly to link this presumed critical period with language change, when he proposed
that only children can carry out major linguistic alterations, which “optimize” their grammars:

due to deterioration or loss in the adult of the ability to construct optimal (simplest) grammars
on the basis of a restricted corpus of examples … I conjecture that changes in later life are
restricted to the addition of a few rules in the grammar and the elimination of rules and hence
a wholesale restructuring of his grammar is beyond the capabilities of the average adult.

(ibid.: 74).

Halle's ideas were taken up particularly by those working within the transformational-generativist paradigm:
“Simplification typically occurs in the learning of speech by children” (Kiparsky 1968: 195); “A child rarely, if
ever, constructs a grammar more complex than that of his models” (King 1969: 74).

Those continuing this tradition now argue that parameter setting by children (the selection and fixing of
pre-set options) is the cause of major changes: “If one aims to understand language change partly in terms
of the way languages are acquired by young children, obsolescence must be treated as a by-product of
some new parameter setting” (Lightfoot 1991: x).

However, this view begs a number of questions. Most radically, is there any evidence that language change
truly occurs between generations? The question turns out to be oversimple: children's language acquisition
alters in character across the years, with youngsters tuned in to different aspects of language at different
ages. They are also subjected to changing social pressures. So at the very least, the question needs to be
examined in agebands (Kerswill 1996).

Children are highly sensitive to phonetics/phonology in their first few years. Typically, youngsters pick up
female-dominated sound changes, judging from work on Philadelphia English: children aged 3 and 4 “are
receptive to the dialect influence of their caregivers at a time when the caregivers are most likely to be
female and locally based” (Roberts 1997: 264), with the result that “it is the female-dominated sound
changes that are advanced in early language learning” (ibid.: 264) - though males and older children may
also play a role (Kerswill 1996). The generation gap is not readily apparent early in life, therefore. Children
match their speech to those around (Roberts and Labov 1995). At the most, children who pick up on a
phonetic change in progress may well advance it further as they get older. Child overgeneralizations (such as
foots for ‘feet’) fade away, and non-standard forms found in child language are mostly unlike those found
in language change. So imperfect learning by youngsters is possibly a mirage (Vihman 1980; Bybee and
Slobin 1982: Labov 1989a; Aitchison 1981).

Schoolchildren aged 6–12 gradually move away from parental influence, and become progressively affected
by their contemporaries: “At the preadolescent stage, we find the beginnings of a move from parent-oriented
to peer-oriented networks” (Kerswill 1996: 196). The notion that children might have “optimal grammars” is
unsubstantiated. More probably, children, like adults, set up alternative analyses, which may or may not get
eventually decided upon (Hankamer 1977; Guy and Boyd 1990) - though an eventual decision might have
longer-term consequences, in that it could be regarded as a trigger for parameter setting, or, in more
traditional terms, the starting point of a reanalysis with (eventually) far-reaching consequences.

No child language event happens sufficiently fast or thoroughly for a parameter to be set or reset in one
swoop, however one identities the various parameters. In the short term, changes tend to be small-scale and
“local” (in the sense of Joseph 1992) within both child language and historical linguistics. The notion of
parameter setting may therefore be useful primarily as a metaphor encapsulating the long-term need for
pattern tidying and pattern maintenance.

12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 3 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

A generation gap develops mainly when children identify strongly with peer groups, which commonly
happens in adolescence (e.g., Bailey and Maynor 1988; Cheshire 1982; Eckert 1988; Romaine 1984). No
absolute “critical period” cut-off point is found for language acquisition. Instead, a gradual decline occurs
(Newport 1991) - though this does not include the lexicon, which is expanded throughout a person's life. In
their teenage years, children's vocabulary escalates, and often diverges from that of their parents. In English,
a leap in vocabulary size around the age of 14 is associated with the acquisition of rules for word derivation
(Aitchison 2000). Teenagers also nurture their “own” lexical items, which tend to accelerate language
change. New coinages acquire regular endings by default, as with blagged ‘conned, exaggerated’ (Ayto
1990); so do extended uses of existing words, as with shooted up ‘injected’ (of drugs). This hastens the loss
(in English) of irregular past tenses and plurals, especially as children may rarely hear old forms such as gelt,
once the past tense of geld.

This does not happen only in England. In Papua New Guinea, Tok Pisin is a pidgin/creole whose lexicon is
based heavily on English, and whose syntax is a mix of English and local languages. First-generation Tok
Pisin creole speakers (those who have acquired pidgin Tok Pisin as a first language) show a marked
divergence from their pidgin-speaking parents. Consider their treatment of plurals (Aitchison 1990).
Teenage speakers have inherited a Tok Pisin plural prefix ol, as with ol man na ol meri ‘men and women,’
versus singular man na meri ‘a man and a woman.’ Many also speak English, and have imported English
plural -s for use with words which are either borrowed from English, or in which there is a strong similarity
between the Tok Pisin and English lexical items. This happens above all in three areas: loanwords for food,
as sandwiches, drinks, tsips ‘chips’; words for periods of time, as minits, auas ‘hours’, wiks, wikends,
aftenuns; and words for people, as frens ‘friends,’ bratas ‘brothers,’ sistas. But this is not just an English
take-over, because in numerous cases both plurals are used, as ol sandwiches, ol frens. Any plural tends to
be preceded by either ol or a numeral, as tri wiks, whether or not an English -s is attached. How the
situation will be resolved in the long run is unclear, though double marking in these three vocabulary areas
is creeping into others, as with ol stons ‘stones.’ These teenagers are therefore making use of a new type of
plural, one not found in the speech of their older relatives.

To summarize, babies do not initiate changes. Groups of interacting speakers do, particularly adolescents.
Any permanent change happens largely via the vocabulary. Change also happens when casual styles of
speech become accepted in more formal settings. Occasional accounts of communities which showed a leap
between generations are possibly due to a failure to record the full range of stylistic alternatives: the gap
between old and young in Fox (an American Indian language), for example, turned out to be due to a
preference for formal styles among the old, and informal among the young, though both styles were
available to each group (Goddard 1989).

But why have so many intelligent linguists been prepared to adopt the “babies rule” viewpoint? Are they
simply attempting to make language history “fit into … the hottest Designer Models” (Lass 1997: xiii)?
Traditions within the subject are largely to blame, it turns out: false models of change were instilled into
generations of linguists. A “tadpole-to-frog” model (my term) was widespread until the 1960s: a linguistic
tadpole was assumed to gradually “turn into” a later frog. How this happened was a puzzle to many, and a
“change-between-generations” model was readily adopted.

Sociolinguists, and particularly William Labov, have now solved the “tadpole-to-frog” problem, by proposing
a “young-cuckoo” model of change (my term): a new form arises in competition with the old, then increases
in use, and finally, like a young cuckoo, pushes the older form out of the nest. A number of historical
linguists have subsequently adopted and popularized such a competition model (e.g., Kroch 1989a). A
“multiple-births” model (my term) is gradually replacing the young-cuckoo one (Aitchison 1995): often,
multiple variants exist in different parts of a community, or in different styles. They fluctuate for a time,
maybe even for generations. Gradually, one variant wins out over the others.

In the case of children and change, then, close-grained sociolinguistic studies have shown that some
proposed psycholinguistic explanations are a mirage. But this does not imply that all
psycholinguistic/cognitive explanations are irrelevant, as will be discussed below.

2 Speech Processing

Speech processing mechanisms are likely to cause change, it has been argued, though in practice speech
comprehension has been invoked more often than speech production.

12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 4 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

Two basic problems underlie discussions of this issue. The first is the long time lag between any supposed
processing need and its final psycholinguistic effect. A second problem (as noted earlier) is how to draw the
line between psychological explanations and other types (linguistic, pragmatic, typological, etc.). For
example, verb-object closeness is statistically the norm (Tomlin 1986). Is the historical insertion of an object
for a finite verb which lacks one (e.g., Joseph 1980b) a linguistic explanation or a psycholinguistic one? We
could either say “finite verbs need objects” (linguistic) or “the human mind perceives certain actions as being
performed on something” (pragmatic/psychological). Unfortunately, linguists themselves vary as to whether
they label such an explanation psycholinguistic or not.

The following account therefore takes into consideration only those papers where the authors have
themselves claimed a psycholinguistic explanation for change. It omits topics such as “transparency” (e.g.,
Lightfoot 1979), since this falls into linguistic (level two) explanations of pattern maintenance. This account
also leaves out non-historically oriented cognitive discussions of discourse (e.g., Givón 1990, who talks
about “the cognitive basis of discourse coherence”) and linguistic availability (e.g., Keenan and Comrie's
1977 “accessibility hierarchy”). It also passes over the multiple publications relating to cognition and
language origin (see Aitchison 1996 for a summary), and this inevitably also leaves aside the copious recent
literature on grammaticalization (e.g., Hopper and Traugott 1993; Fischer 1997; Heine and Bybee, both this
volume).

Most of the work in this (now narrowly defined) area has assumed that comprehension (parsing) needs have
affected linguistic structure. These claims have varied from decade to decade, and will be dealt with below in
roughly chronological order.

An imbalance between the needs of memory and those of perception might cause change, according to an
early influential paper (Bever and Langendoen 1972). The loss of English noun inflections between the
eleventh and fifteenth centuries was due to the complexity of these inflections, which put too great a strain
on the memory, Bever and Langendoen argued. But after the loss of inflections, it became perceptually
difficult to split sentences up into main and subordinate clauses when a relative clause was involved, because
the relative pronoun did not have to be included, as in:

(2) Lete fetche the best hors maye be founde (Mallory, fifteenth century)

This processing difficulty, they argue, led to the relative pronoun becoming obligatory in sentences of this
type.

Bever et al. (1977) also argued that processing mechanisms can cause the acceptance of strictly
ungrammatical sentences, as:

(2) One of my elephant's birthdays is today (my example)

This is technically ungrammatical, and should be:

(3) One of my elephant's birthday is today

which most people regard as odd. Bever and his colleagues were therefore among the earliest authors to
argue for processing explanations of language change.

A problem with this whole field is the one-off nature of various interesting speculations. For example, Menn
and MacWhinney (1984) pointed out that many languages avoid repeated morphs, and proposed a
processing (perceptual) explanation. For example, un- cannot be attached to a word already beginning with
that prefix, so *an un-unhappy man becomes a not unhappy man (Aitchison and Bailey 1979).

An important set of explanations in the 1970s and 1980s related to word order and word order change.
Kuno (1974) was an early paper which argued that relative clauses preceded nouns in OV languages, but
followed them in VO ones in order to avoid center-embedding, which is perceptually difficult to process. The
paper was, however, more psycholinguistic than historical, in that it did not specify how changes toward this
configuration might come about. Nor did it explain why the majority of the world's languages, including
many OV ones, have postnominal relative clauses, a matter taken up by Antinucci et al. (1979). This claimed

12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 5 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

many OV ones, have postnominal relative clauses, a matter taken up by Antinucci et al. (1979). This claimed
that prenominal relatives cause perceptual problems: matrix and subordinate clauses cannot be reliably
distinguished, and so they tend to be avoided.

But John Hawkins was perhaps the first to try seriously to link psychological/ typological findings with
language change, in a series of books and papers stretching across decades (e.g., Hawkins 1979, 1983,
1988a, 1988b, 1994). Most notably, he put forward a principle of cross-category harmony, which
coincidentally tied in with X-bar theory. This principle asserted that “there is a quantifiable preference for
the ratio of preposed to postposed operators within one phrasal category … to generalise to others” (1983:
134). He realized the historical difficulty involved in this: if languages are striving toward cross-category
harmony, why do they then not achieve their goal? Why do some become inconsistent and change their word
order? He therefore attempted to specify mechanisms by which any changes come about. In this explanation,
he realized the importance of doublets (e.g., John's the hat, the hat of John) which alter their relative
frequencies, and proposed outline constraints for handling this (still not adequately solved) problem.

Hawkins's later work has less to say about how any changes are implemented. He has pointed out, for
example, that the languages of the world overall prefer suffixes to prefixes, whatever their word order type
(Hawkins and Cutler 1988; Hawkins and Gilligan 1988), for which he provides a processing explanation - a
clarity principle, though he has little to say about the historical mechanisms by which this principle is
implemented. In short, he has moved from proposing mechanisms for change to more general claims about
why languages are the shape they are, which he relates to processing needs. In his more recent work, he
argues for “Early Immediate Constituents” (1994): the human parser, he suggests, prefers orders in which
any comprehender can quickly establish the immediate constituents for any given phrase. This accounts for
a wide range of well-known phenomena, such as heavy NP shift. The parser will dislike:

(4) I gave [the huge octopus that was extremely difficult to catch] to Aloysius (my example)

and will prefer:

(5) I gave to Aloysius [the huge octopus that was extremely difficult to catch]

- though other explanations are possible. Wasow (1997) points out that “endweight,” the tendency to place
long complex phrases at the end of sentences, is typically attributed to parsing needs. But the demands of
planning must also be taken into consideration, he argues, and might be paramount.

A general tendency, therefore, seen both in Hawkins's performance theory of order and constituency (1994)
and in linguistic theory in general, is the search for broad-ranging principles which contribute to an
understanding of linguistic ability. However, it would be helpful for those interested in historical linguistics if
the working of such principles could be tied in more directly to specific changes, especially as most changes
start out as “local” ones which only later generalize to a wider set of data (Joseph 1992).

An important aspect of Hawkins's work for historical linguists is that he has always attempted to quantify his
data. Increasingly, linguists are realizing that sociolinguistic variation is not the only sort of data which
needs to quantified: all constraints are potentially violable, and quantification is needed in this area also.
Optimality Theory proposes that each language has its own ranking of these constraints (Archangeli and
Langendoen 1997), and, usefully, work within this framework has started to appear on historical topics,
mostly so far on phonology (Nagy and Reynolds 1997; Zubritskaya 1997; introduction to this volume, n.
135). Perhaps when historical linguists have reliable, quantified typological data on preferred constructions,
they can hypothesize the parsing/ production principles which favor these. Equally importantly, they may be
able to specify how languages alter their constructions in order to fit in with such requirements, and also
explain why they sometimes appear to resist doing so.

Within psycholinguistics, a further massive amount of effort is being put into work on connectionism (parallel
distributed processing). Promising areas are how children acquire past tenses, and how humans identify
words (e.g., Kim et al. 1994; Elman 1993). But so far, only occasional link-ups have been made with
language change (e.g., Tabor 1993), even though the notion of changing connection strengths has potential
implications for this field. However, an interesting recent speculation is that optimality theory and
connectionism might prove to be compatible (unpublished paper by Mark Liberman, quoted in Nagy and
Reynolds 1997).

12/11/2007 03:44 PM

25. Psycholinguistic Perspectives on Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 6 of 6

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747927

Reynolds 1997).

But perhaps the most promising area of language and mind is the rapidly increasing literature on why
languages are the shape they are, and which constructions are liable to change into which others. This work
ties in with broader considerations, such as the origin of language, which is still felt by many historical
linguists to lie outside their field, even though both historical linguistics and language origin studies share
central topics such as grammaticalization. The overall unification of all these burgeoning strands is a hope
for the future.

Cite this article

AITCHISON, JEAN. "Psycholinguistic Perspectives on Language Change." The Handbook of Historical Linguistics.
Joseph, Brian D. and Richard D. Janda (eds). Blackwell Publishing, 2004. Blackwell Reference Online. 11 December
2007 <http://www.blackwellreference.com/subscriber/tocnode?
id=g9781405127479_chunk_g978140512747927>

Bibliographic Details

The Handbook of Historical Linguistics

Edited by: Brian D. Joseph And Richard D. Janda
eISBN: 9781405127479
Print publication date: 2004