John Benjamins Publishing Company
This is a contribution from Linguistic Approaches to Bilingualism 6:6
© 2016. John Benjamins Publishing Company
This electronic file may not be altered in any way.
The author(s) of this article is/are permitted to use this PDF file to generate printed copies to be
used by way of offprints, for their personal use only.
Permission is granted by the publishers to post this file on a closed server which is accessible
only to members (students and faculty) of the author’s/s’ institute. It is not permitted to post
this PDF on the internet, or to share it on sites such as Mendeley, ResearchGate, Academia.edu.
Please see our rights policy on https://benjamins.com/content/customers/rights
For any other use of this material prior written permission should be obtained from the
publishers or through the Copyright Clearance Center (for USA: www.copyright.com).
Please contact [email protected] or consult our website: www.benjamins.com
Codeswitching and the timing of
Lexical Insertion
Jeff MacSwan
University of Maryland
Lillo-Martin, Müller de Quadros and Chen Pichler (2016) (henceforth LQC) apply a Constraint-free (CF) approach to codeswitching (CS) (MacSwan, 2014) to
bimodal language mixing; they pursue an account which would show that, like
spoken bilingual language mixing, language mixing across modalities of speech
and sign may be analyzed without the need for CS-specific constraints (as in a
Constraint-based, or CB, approach). Their specific choice of model calls for a
Minimalist Approach augmented by Distributed Morphology (DM). The tension
between this approach and classical Minimalism relates to the timing of lexical insertion. Chomsky (1995, 1998) takes what Chametzky (2003) calls a “lexical-entry
driven” (or Lexicalist) approach for granted: Lexical Items (LIs) are selected into
a Lexical Array, where Merge, Agree and perhaps other syntactic operations build
and transform structure in a way that responds to lexically specified properties.
DM postpones Vocabulary Insertion until the end of the derivation, as in earlier
generative models. DM implementations of a CF approach to CS (LQC; BandiRao, & den Dikken, 2014; Grimstad, Lohndal, & Åfarli, 2014) deserve serious investigation, but present some challenges for codeswitching data which are readily
overcome by an early insertion version of classical Minimalism. To contextualize
these challenges, we review some of the limitations of earlier late lexical insertion
models of CS before narrowly focusing on challenges which might remain for a
DM solution to a CF approach to CS, including LQC’s Language Synthesis Model.
Sankoff and Poplack (1981) provided important early insights into the formal problem associated with lexical insertion and CS. Their study began with the
simplest assumption, postulating that a bilingual grammar comprises the simple
union of two phrase structure grammars, and nothing more – a CF approach.
Working with Spanish-English CS data, a “free union grammar”, as they termed
it, would include all of the phrase structure rules common to both English and
Spanish, as well as those that differed, such as NP → Det Adj N for English and NP
Linguistic Approaches to Bilingualism 6:6 (2016), 786–791. doi 10.1075/lab.6.6.08mac
issn 1879–9264 / e-issn 1879–9272 © John Benjamins Publishing Company
Codeswitching and the timing of Lexical Insertion 787
→ Det N Adj for Spanish. Lexical insertion rules associated terminal nodes with
lexical items(LIs) (e.g., Det → the, N → house, Adj → while). Sankoff and Poplack
observed that the free union of the two phrase structure grammars would result in
ill-formed constructions; not only would it generate ill-formed codeswitches, like
the casa white (‘the white house’), but it would generate ill-formed monolingual
constructions too, such as the house white, where English lexical insertion rules
apply to terminal nodes arranged by Spanish phrase structure rules.
Although Sankoff and Poplack (1981) expressed a preference for avoiding CSspecific mechanisms to mediate between the two grammars, they concluded that
such a mechanism is necessary on empirical grounds. In order to constrain the
grammars so that they do not generate ill-formed expressions, Sankoff and Poplack
introduced a superscripting mechanism called a language tag which restricted lexical insertion rules so that the grammar contributing the phrase structure rule
would also be the grammar from which lexical insertion rules were drawn. Hence,
under conditions of CS, the Spanish phrase structure rule would be annotated, as
in NP → Det Nsp:n Adjsp:adj, giving us the casa blanca. The fundamental problem
Sankoff and Poplack encountered in their late insertion model was that the phrase
structure rules defined the word order before lexical insertion. This created problems for a bilingual grammar, because the grammar made commitments to word
order before inserting words, such that a grammar would permit the structure of
one language or the other regardless of how the terminal nodes were lexically filled.
Sankoff and Poplack’s language tagging system constituted a Constraint-based
Approach because it recruited ad hoc mechanisms that were not, and could not be,
independently motivated to account for either Spanish or English. Furthermore,
the language tag, which resurfaced in GB-oriented approaches as a language index
(Di Sciullo, Muysken, & Singh, 1986) and a language feature (Belazi, Rubin, &
Toribio, 1994), introduced an ordering paradox: A language L is the output of a
grammar G, so L cannot be any part of G.
These issues disappear in a Lexicalist Approach. Because lexical insertion takes
place at the outset, we may conceptualize a CF approach to CS as the free union of
the two grammars, {Gx ∪ Gy}, so that the facts of language mixing are explained
in terms of principles and requirements of the specific grammars in interaction,
and nothing more. Select draws LIs from discretely represented Lexicons1 into a
Lexical Array, where Merge and other operations continue the mapping to PF and
LF. Word order falls out of syntactically-relevant properties of the LIs selected into
the Lexical Array. Figure 1 offers a graphic representation of a bilingual version of
this model (MacSwan, 2016).
1. Of course, the discreteness of bilingual representations cannot be taken for granted. For a
review of arguments motivating this claim, see MacSwan and Colina (2014).
© 2016. John Benjamins Publishing Company
All rights reserved
788 Jeff MacSwan
To avoid pitfalls of prior approaches, three important conditions must be met
by any DM-inspired late insertion CF approach to CS: (1) the feature matrices of
LIs must remain sufficiently rich post-syntactically (after the deletion of formal
features) to ensure that word order and lexical insertion align in the way anticipated by the data, avoiding the trouble Sankoff and Poplack (1981) encountered; (2)
the feature specification of LIs (or VIs, Vocabulary Items) must be independently
motivated – not motivated by the CS facts alone – to avoid the use of language
tag proxies; and (3) proposals must be made fully explicit, detailing the nature of
specific features, as vague reference to contrasting “feature bundles” may inadvertently serve as proxies for language labels.
Lexicon
(Most Affixation,
Lexical Phonology)
Lexicon
(Most Affixation,
Lexical Phonology)
Select
Cycle again at each Phase
Lexical Array
(Internal, External)
Merge
(Multiple) Spell-Out
Phonological
Component
PF (PHON)
Covert
Component
LF (SEM)
Figure 1. A minimalist model of bilingual codeswitching (MacSwan, 2016)
With these challenges in mind, let us consider a couple of CS examples. Consider
the pronominal CS data below, due to van Gelderen and MacSwan (2008).
Codeswitching between a subject pronoun and a verb is ill-formed, but switching between a lexical subject and a verb is not, as shown in (1) and (2). What is
the specific array of features making up Spanish pronouns and English pronouns
at the time of lexical insertion such that insertion is guided to avoid CS in (1)? In
(3), under conditions of conjunction, the structure is well-formed for CS, meaning
that lexical insertion is once again allowed. How do we account for these contrasts
in a late insertion model?
© 2016. John Benjamins Publishing Company
All rights reserved
Codeswitching and the timing of Lexical Insertion 789
(1) a. *Yo fightall thetime
I fight allthe time
b. *Ellos fightall thetime
they fight allthe time
c. *Ella fights all thetime
she fights allthe time
(2) a.
b.
Mi novia
fights all thetime
my girlfriend fights allthe time
Mis amigos fightall thetime
my friendsfight allthe time
(3) a.
b.
c.
d.
Mis amigos y yo fightall thetime
my friendsand I fight allthe time
Yo y mis amigos fightall thetime
I and my friendsfight allthe time
Mis amigos y el fightall thetime
my friendsand he fight allthe time
Mis amigos y ellos fightall thetime
my friendsand theyfight allthe time
Consider another example, due to Mahootian (1993). Farsi, an OV language, and
English, a VO language, generate OV word order in CS if the verb is Farsi, regardless of the language of the object; and VO word order if the verb is English – again,
regardless of the language of the verb.
(4) a.
b.
Ten dollars dad-e
ten dollars give-perf
‘She gave ten dollars.’
jaedid when you sell your own house
Tell them you’ll buy xune-ye
Tell them you’ll buy house-poss new when you sell your own house
‘Tell them you’ll buy a new house when you sell your own house.’
A standard account of VO/OV word order attributes verb movement to properties of T, prompting raising in the case of English but not in the case of Farsi,
deleting formal features along the way. Once the structure is formed by the introduction of formal features, what prevents an English verb from being inserted in (4a) rather than a Farsi verb? What are the specific features of Farsi and
English verbs at Vocabulary Insertion that allow the bilingual grammar to select
the appropriate form?
Many other examples could be offered, but these two suffice to make the point.
CS exhibits specific word order characteristics and grammaticality effects which
critically depend on the grammar having a mechanism to align structure and
© 2016. John Benjamins Publishing Company
All rights reserved
790 Jeff MacSwan
lexical insertion. In a Lexical Approach, LIs are inserted into the derivation before
structure is built; in DM, vocabulary must find its way to the right node, guided by
a richly specified feature array. To properly motivate a DM account of CS within
a CF approach, considerably more work must be done, showing that such an approach can attend to old and new puzzles alike. Certainly, it is worth the effort
to do so, and LQC’s article proposes an outline of such an approach; but it leaves
many questions unanswered.
References
Bandi-Rao, S., & den Dikken, M. (2014). Light Switches: On v as a pivot in codeswitching, and
the nature of the ban on word-internal switches. In J. MacSwan (Ed.), Grammatical Theory
and Bilingual Codeswitching (pp. 161–184). Cambridge: MIT Press.
Belazi, H. M., Rubin, E. J., & Toribio, A. J. (1994). Code switching and X-bar theory: The functional head constraint. Linguistic Inquiry, 25(2), 221–237.
Chametzky, R. A. (2003). Phrase structure. In R. Hendrick (Ed.), Minimalist Syntax (pp. 192–
225). Oxford: Blackwell. ​doi: 10.1002/9780470758342.ch5
Chomsky, N. (1995). The Minimalist Program. Cambridge: MIT Press.
Chomsky, N. (1998). Minimalist inquiries: the framework. MIT OPL 15. Dept. of Linguistics,
MIT.
Di Sciullo, A.-M., Muysken, P., & Singh, R. (1986). Government and code-switching. Journal of
Linguistics, 22, 1–24. ​doi: 10.1017/S0022226700010537
Grimstad, M. B., Lohndal, T., & Åfarli, T. A. (2014). Language mixing and exoskeletal theory: A
case study of word-internal mixing in American Norwegian. Nordlyd, 41(2), 213–237.​
doi: 10.7557/12.3413
Lillo-Martin, D., Müller de Quadros, R., & Chen Pichler, D. (2016). The development of bimodal bilingualism: Implications for linguistic theory. Linguistic Approaches to Bilingualism,
6(6), 719–755. ​doi: 10.1075/lab.6.6.01lil
MacSwan, J., & Colina. S. (2014). Some consequences of language design: codeswitching and
the PF interface. In J. MacSwan (Ed.), Grammatical Theory and Bilingual Codeswitching (pp.
185–200). Cambridge: MIT Press.
MacSwan, J. (2014). Programs and proposals in codeswitching research: Unconstraining theories of bilingual language mixing. In J. MacSwan (Ed.), Grammatical Theory and Bilingual
Codeswitching (p. 1–34). Cambridge: MIT Press.
MacSwan, J. (2016). Codeswitching in adulthood. In E. Nicoladis, & S. Montanari (Eds.), Lifespan
perspectives on bilingualism (pp. 183–200). Washington, DC: American Psychological
Association & Berlin: Walter de Gruyter.
Mahootian, S. (1993). A null theory of code switching. PhD dissertation, Northwestern University.
Sankoff, D., & Poplack, S. (1981). A formal grammar for code-switching. Papers in Linguistics,
14, 3–45. ​doi: 10.1080/08351818109370523
van Gelderen, E., & MacSwan, J. (2008). Interface conditions and codeswitching: An F-movement
analysis of pronouns and lexical DPs. Lingua, 118(6), 765–776.​
doi: 10.1016/j.lingua.2007.05.003
© 2016. John Benjamins Publishing Company
All rights reserved
Codeswitching and the timing of Lexical Insertion 791
Author’s address
Jeff MacSwan
University of Maryland
3942 Campus Drive, Room 2223
College Park, MD 20742
[email protected]
© 2016. John Benjamins Publishing Company
All rights reserved