The departure from the Unmarked Alignment Hypothesis is marked by the first deviation from the canonical sentence schema. Empirical studies of the second language acquisition of a range of configurational languages, such as German or English, identified sentence-initial adverbials and WH-words to be the first non-subjects to occur in sentence-initial position in L2 development. In this section we will first consider these post-initial L2 developmental dynamics in English and German, and we will then attempt to represent the underlying mechanisms in more general terms.
In Section 3.2 we described roughly how structures such as sentence-initial adverbials and WH-words can be accounted for. Now it will be useful to see how non-linearity is created in these structures using the correspondence mechanisms sketched out above. The mechanism applying to sentence-initial adverbials was briefly discussed earlier. It stipulates that constituents adjoined to XP be assigned one of the non-argument functions TOP, FOC or ADJUNCT. This licences sentence-initial arguments adjoined to XP to assume one of these grammatical functions (Bresnan, 2001). One principle governing c- to- f-structure mapping stipulates that specifiers of functional projections are grammaticalised discourse markers (i.e. TOP, FOC or SUBJ). Example (42), which repeats (5), illustrates that in the case of English, the specifier of IP is SUBJ and that the constituent adjoined to XP is ADJ.
As mentioned above, in (42) the initial position is occupied by a non-subject. This marks a departure from the Unmarked Alignment Hypothesis, which assumes that the three parallel levels of syntax are mapped onto each other in a strictly one-to-one manner, thus defining the first sentential position as the default for NPSUBJ. The mapping of AP onto ADJUNCT and of DP onto SUBJ is now no longer linear. Instead, assigning the grammatical function ADJ to AP is based on XP-adjunction, and
assigning SUBJ to DP now relies on identifying the specifier of functional projections in c-structure. Note that at this point in interlanguage development the rest of the canonical pattern can nevertheless be mapped one-to-one from c-structure onto the hierarchy of grammatical functions.
Figure 8: XP-adjunction in interlanguage
The assumption that in the presence of XP-adjunction the rest of the canonical pattern can nevertheless be accounted for by one-to-one mapping is supported by the developmental trajectories found in German, Swedish and English interlanguage systems. Note that in native German XP-adjunction constrains the verb into second position (cf. Berman and Frank, 1996; Berman, 2003). This is similar in Swedish. In English, the XP-adjunction of WH-words that refer to non-subjects constrains an auxiliary into second position (cf. Kaplan and Bresnan, 1982). It is a well-attested finding from research on the acquisition of German as a second language (GSL) that learners of GSL always violate this constraint when they first acquire XP-adjunction (cf. Clahsen, Meisel and Pienemann, 1983; Pienemann, 1981, 1998a). The same is true for Swedish as a second language (cf. Pienemann and Håkansson, 1999; Håkansson, Pienemann and Sayehli, 2002). In a similar vein, ESL learners initially form WH- questions without the auxiliary in second position (cf. Pienemann, 1998a), thus applying
canonical word order after XP-adjunction. Note that all of these interlanguage rules are obligatory.
Pienemann (1998a) showed that the operations required to produce the verb-second constraint in the above Germanic languages are based on the transfer of grammatical information in the S-procedure, and this is an operation that occurs much later in L2 development than XP-adjunction. Hence, for an extended period in L2 development, learners of German, Swedish and English as L2 produce c-structure configurations that violate the verb-second constraint of the target language. Of course, in English the verb- second constraint applies to sentences with WH-words (referring to non-subjects) in XP only.
The interlanguage violation of the verb-second constraint is exemplified in (43). This example is taken from a longitudinal study of an eight-year-old Italian girl acquiring German as L2 - after about one year of contact in a natural setting.
(43) auf ein blatt wir schreiben was die sagt
on a sheet (of paper) we write what she says ―we write on a sheet of paper what she says‖
(Eva, week 56, Pienemann 1981: 58)
Native German would have required the verb to be placed in second position as shown in (44).
(44) Auf ein Blatt schreiben wir, was sie sagt.
―on a sheet of paper write we what she says.‖
As this example illustrates, at this stage the c-structure produced by L2 learners can be accounted for by two principles: (i) XP-adjunction and (ii) direct mapping.
In this context it is worth noting the following empirical facts that strongly support this assumption of a greatly simplified interlanguage rule system. In their survey of more than one thousand informants Pienemann and Håkansson (1999) found this interlanguage rule to be categorical. The same is true for the extensive longitudinal and cross-sectional studies by Pienemann (1981, 1998a) which did not document even a
single instance of a verb-second structure at this point in development. In addition, Håkansson, Pienemann and Sayheli (2002) found that even Swedish learners of German follow this developmental trajectory, although the verb-second constraint is present in both languages. Pienemann, Di Biase, Kawaguchi and Håkansson (this volume) explain this on the basis of the ―developmentally moderated transfer hypothesis.‖ In other words, the interlanguage state described by the above set of two rules is supported by robust data.
It is important to note that the changes to the interlanguage system described above are brought about exclusively by XP-adjunction. In other words, these changes can be described solely in terms of the relationship between c-structure and f-structure. The correspondence between a-structure and f-structure remains unaffected by this.
One aspect of XP-adjunction is somewhat similar to processes found in English and German WH-questions. As mentioned above, the c-structure shown in Figure 8 is based on the following simplified rule (cf. Dalrymple, 2001, 406):
(45) CP → XP C‘
(↑FOCUS)=↓ ↑=↓ (↑FOCUS)= (↑COMP* GF)
In order to secure completeness and coherence we must assume that the discourse function FOC is allowed to satify the unsatisfied argument function (i.e. OBJ), as illustrated in the f-structure in Figure 9 where the FOC function is linked to the OBJ function. As can be seen in Figure 9, the WH-question constituent maps onto both, FOCUS and OBJ functions.
In other words, information about the link between FOCUS and OBJ needs to be exchanged between the two grammatical functions, and this information exchange constitutes one aspect of non-linearity that is present in WH-questions. An additional aspect of non-linearity is created by the fact that the assignment of SUBJ to a constituent is no longer canonical and instead relies on identifying c-structure regularities similar to the XP-adjunction of ADJUNCTs that was discussed above.
As we have seen, WH-questions and XP-adjunction have in common that a constituent annotated for a function other than SUBJ appears in initial position. In English, the difference between the two is that XP-adjunction has no further repercussions in c- structure,30 whereas in English WH-questions the auxiliary is constrained to appear in second position. This can be achieved by tensed auxiliaries appearing in C when inversion is involved and in I when it is not. Dalrymple (2001, 64) adds that the auxiliary appears in the proper position on the basis of ―[c]onstraints on the functional structure of constructions requiring or forbidding subject-auxiliary inversion…‖ Bresnan (1982) and Pinker (1984) specified such constraints on the functional structure in the original version of LFG to account for the position of English auxiliaries.
Pienemann (1998a) showed that the constraints on the functional structure specified by Bresnan (1982) and Pinker (1984) require the type of transfer of grammatical information that is possible only on the basis of information transfer in the S-procedure. The presence of a constituent annotated for FOCUS, however, does not require such information transfer. The non-linearity of this structure is limited to a discourse function appearing in the default position of SUBJ. We have shown for XP-adjunction that the remainder of the mapping process may follow a canonical pattern – as illustrated in Figure 7.
In other languages (e.g. in Finnish 31) WH-question formation has no further repercussions for c-structure, not only if the WH-word is the grammatical subject, but also for other grammatical functions. Therefore these structures can be produced by learners as soon as the WH-word can occupy the first position in the sentence.
30
This is not the case in German, which requires the inflected verb to appear in second position (cf. Berman and Frank, 1996).
31
Table 1: German L2 development in c- to f-structure mapping and constraints on c- structure
The developmental pattern that has emerged so far for German SLA is summarised in Table 1, which lists the structures discussed above in their developmental order, together with a number of key features. Table 1 shows that XP-adjunction is acquired as one of the cumulative developmental features of the interlanguage. WH-questions are first formed with non-core arguments in XP position and later with core arguments. Up to this point (i.e. Stage 3) interlanguage structures can be produced with XP-adjunction and direct mapping. The latter, however, must be abandoned when in WH-questions the verb appears (correctly) in second position, based on (target-like) constraints on the functional structure of these constructions, because in the resulting c-structure the grammatical function of a core argument in XP position can only be identified correctly if there is a link between its core function and its discourse function in f-structure - as shown in Figure 9. It is this link that creates a degree of linguistic non-linearity that places this structure at the top of the hierarchy in Table 1.
At a more general level, a number of basic principles emerge from the above overview of German L2 development. It starts out with the Unmarked Alignment Hypothesis, which is characterised by complete adherence to a linear correspondence relationship in which the first and most prominent position in c-structure is occupied by the most prominent syntactic function (the SUBJECT), representing the most prominent argument available. The appearance of an AP in this position constitutes the first modification of this linear correspondence. At this point, XP-adjunction permits constituents to be marked for discourse functions, while the rest of c-structure is
mapped canonically onto the universal hierarchy of grammatical core functions. This state of the interlanguage, with XP adjunction and canonical mapping, is attested as the stage following the Unmarked Alignment Hypothesis in the developmental trajectories of a large number of second languages, including Japanese (see Kawaguchi, this volume), Italian (Di Biase and Kawaguchi, 2002), Spanish (Taylor, 2004) and Turkish (Özdemir, 2004).
XP-adjunction triggers the differentiation of the syntacticised discourse functions TOPIC and FOCUS32 from SUBJECT in the developing interlanguage system. Direct mapping at Level 2 does not allow for the differentiation of SUBJECT and TOPIC. Instead, if SUBJ is present it will always occupy the first position. This close connection between SUBJECT and TOPIC is also reflected in Bresnan‘s typological perspective. She states that it ―…comes from the universal default that optionally identifies SUBJ and TOP‖ (Bresnan, 2001, 117). The equilibrium of the direct mapping processes is disturbed once adjuncts or WH-words appear in the default position of the constituent mapped onto SUBJ. This was shown in Figure 8 above. When the constituent under XP is mapped onto a discourse function (as in WH-questions), TOPIC is differentiated from SUBJECT
The mapping principles and their structural outcomes are summarised in Figure 10. To account for these dynamics we propose the TOPIC hypothesis in (46).
(46) The TOPIC hypothesis.
In second language acquisition learners will initially not differentiate between SUBJ and TOP. The addition of an XP to a canonical string will trigger a differentiation of TOP and SUBJ which first extends to non-arguments and successively to non- arguments, thus causing further structural consequences.
32
Bresnan and Mchombo (1987, 757-8) adopt three principles relating to the role of TOP and FOC functions in the grammars of natural language: 1. the relativised constituent or relative pronoun in relative clauses universally bears the TOP function; 2. the interrogative pronoun or questioned constituent universally bears the FOC function; and 3. the same constituent cannot be both focus and topic of the same level of functional clause structure.
discourse principle c- to f- mapping structural outcomes
Topicalisation of core arguments
TOP= OBJ The TOP function is assigned to a core argument other than SUBJ
↑ ↑ ↑
XP adjunction TOP=ADJ Initial constituent is a circumstantial adjunct or a FOCUS WH-word. TOPIC is differentiated from SUBJECT
↑ ↑ ↑
Canonical Order SUBJ = default TOP
TOPIC and SUBJECT are not differentiated
Figure 10: The Topic Hypothesis
The TOPIC hypothesis intends to capture the development of syntacticised discourse functions in second language acquisition. The actual position of the adjunct (before or after the canonically mapped structure) is a c-structure issue which the learner may resolve either way, depending on the specific c-structure constraints of the L2. For instance, in Japanese, a verb-last language, only pre-verbal positions are possible (cf. Kawaguchi, this volume). Once core-arguments appear in initial position, however, linear correspondence is no longer viable. In other words, it is the dynamics of the developmental process, starting with the use of non-subjects in focus position, that leads to the eventual collapse of the learner‘s exclusive reliance on purely canonical association.