Chicago under revision

Materials
Two texts had been constructed for the corresponding Russian experiment in Thomson (2000) to accompany picture stories taken from Takahashi & Frauman-Prickel (1985). One described a man washing his hands (ten pictures in all). The other described a man becoming sleepy and going to bed (eighteen pictures in all, one excluded, leaving seventeen). The texts were divided into segments corresponding to the pictures, with each segment containing one or more sentences. The texts were modified by the insertion of occasional inflectional anomalies (in one out of three segments). An anomaly in the first segment of the first text was used for training. The placement of anomalous inflectional forms was not deliberately randomized, since the categories of anomalies were decided in advance, and there was a need to find plausible locations for the each type of anomaly. Nevertheless, their distribution was clearly nonsystematic. The texts from the Russian experiement were translated into Kazak by one of the authors (Uataeva) and checked with other native speakers.
It was not obvious how to create syntactically/semantically anomalous verbal inflectional forms in Kazak strictly parallel to those in the Russian texts, and therefore two unrelated anomalous verbal inflectional forms were used (in part to keep the total number of anomalies nearly the same, and in part in an exploratory spirit). One of the verbal anomalies involved person agreement (as was did one anomaly in Experiment 2 of Thomson, 2000). The form il-di, hang-past, 'he hung' (transitive) was replaced with il-di-Niz, hang-past-second.person.polite, 'you (polite) hung'. There was thus a clash between the overt third person subject (adam, 'person'), and the second person polite verbal subject agreement. The other verbal inflectional error involved phonological place harmony. The form tösek-ten, bed-ablative.case, 'from the bed' was replaced with the non-harmonizing tösek-tan. Both of these verbal inflectional anomalies were of interest in their own right and they will be included in the discussion. However, only the case marking anomalies will play a role in the comparison of the results from the Russian experiment with the results from the Kazak experiment.
The Kazak case anomalies were based on the Russian case anomalies, which were as follows:
Oblique-for-nominative replacements:
1. The replacement of required nominative case by anomalous instrumental case (c&elovek, 'person' replaced by c&elovekom).
2. The replacement of required nominative case by anomalous locative case (br'uki, 'pants(pl)' replaced by br'ukax).
Oblique-for-oblique replacements:
3. The replacement of required locative case by anomalous instrumental case (krovati, 'bed' replaced by krovatju).
4. The replacement of required instrumental case by anomalous locative case (mylom, 'soap' replaced by myle).
Nominative-for-oblique replacements
5. The replacement of required locative case by anomalous nominative case (kresle, 'armchair' replaced by kreslo).
6. The replacement of required instrumental case by anomalous nominative case (polotencem, 'towel' replaced by polotenco).
Thus, there were two instances of obliques replacing nominatives, two instances of obliques replacing obliques, and two instances of nominatives replacing obliques. The Kazak version employed parallel case-marking anomalies with one partial exception. The replacement of instrumental by locative was only detected by one third of native listeners in early piloting. This appeared to be due to the fact that the substituted case-marked noun happened to match the initial portion of the immediately following word in the sentence, giving the impression of a simple hesitation and back-track, and thus not creating a case anomaly for most native listeners. Therefore, rather than the locative form, the ablative case form was substituted for the instrumental case form (the meaning would then be 'from the soap' rather than the required 'with the soap'). This still consisted in an oblique-for-oblique subsitution. However, it assumed that sensitivity to oblique forms (and contexts) develops equally for different case values, which is of course, doubtful. Still, in relative terms, we did expect clear processing differences for oblique case forms as opposed to nominative forms (see Thomson, 2000 for further dicussion).
The case substitutions used in the Kazak experiment were thus as follows:
Oblique-for-nominative replacements:
1. The replacement of required nominative case by anomalous instrumental case (adam, 'person' replaced by adammen).
2. The replacement of required nominative case by anomalous locative case (s&albary 'pants(possessed)' replaced by s&albarynda).
Oblique-for-oblique replacements:
3. The replacement of required locative case by anomalous instrumental case (tösekte, 'bed' replaced by tösekpen).
4. The replacement of required instrumental case by anomalous ablative case (sabynmen, 'soap' replaced by sabynnan).
Nominative-for-oblique replacements
5. The replacement of required locative case by anomalous nominative case (kreslada, 'armchair' replaced by kresla).
6. The replacement of required instrumental case by anomalous nominative case (sülgimen, 'towel' replaced by sülgi).
There are theoretically-based expectations regarding the relative levels of sensitivity to all of the anomalies except the one involving place harmony. The most detectable anomalies should be the person agreement anomaly (rich person agreement being semantically concrete, and hence relatively readily acquirable—this was the most detectable type of anomaly in either of the experiments in Thomson, 2000) and the oblique-for-nominative subsitutions (the claim being, as noted above, that the function of oblique case as a signal of non-subject status, i.e. non-focal status in the mental representation of the meaning, develops more readily than other case functions). It was expected that the least detectable anomalies would be the nominative-for-oblique substitutions, since the nominative case was claimed to be the case which least readily acquires any special functions as a processing trigger. Rather, it is treated by the processor as if it is lacking inflection (which in Kazak, unlike Russian, could be argued to be literally true even for native speakers). Finally, the oblique-for-oblique substitutions were expected to be intermediate in detectability between oblique-for-nominative (high detectability) and nominative-for-oblique (low detectability), since the semantic role-marking functions of oblique cases were held to develop more readily than the semantic role-marking function of nominative case, but less readily than the function of oblique marking as an indicator of nonsubjecthood. However, we were also well aware that this pattern did not hold perfectly for the Russian experiment, especially in relation to the instrumental-for-locative substitution (segment 14 of text 2).
The first task in the experiment, the Dual Task, was a divided attention task, with multiple choice picture selection as the distractor task, and anomaly detection as the central experimental task. For the picture-selection sub-task, drawings from Takahashi & Frauman-Prickle (1985) were arranged in rows of three, labelled a, b, and c. There were twenty-seven numbered rows of pictures corresponding to the twenty-seven segments of the tape-recored texts. There was only a five second interval between segments. This time limit was intended to force rapid decisions in the picture selection sub-task, to allow little or no opportunity for metalinguistic reflection, and to keep listeners strongly focused on the meaning. Under these conditions reactions to anomalies are assumed to occur on a relatively automatic, on-line, language-processing basis.
The tape recording for the second task, Listening Only, included the same segments, but with ten second pauses between them. The Listening Only task thus differed from the first task in terms of overall processing demands for at least three reasons: 1) more time was allowed for basic comprehension processes; 2) the texts were already familiar, having been heard before (with visual pictorial support from the picture selection sub-task to further strengthen the memory of the story content); 3) there was no concurrent task dividing participants' attention, and consuming processing resources.¹³ It might be considered a weakness of this design that the same stimuli were used in all three experiments. A more elegant design might have used three separate matched, counterbalanced texts for the three tasks. Reusing the same stimuli in each task meant that participants who detected a substantial portion of the anomalies in an earlier task had less opportunity to demonstrate improvement on later tasks. However, as the experiment was conceived, text repetition was itself intended as a contributor to increased processing ease in going from task to task.
For the Dual Task and the Listening Only task, answer sheets were provided with written instructions. The answer sheets contained of numbered blank spaces corresponding to each segment of the text. (In the tape recordings, the number of each text segment was spoken immediately before the segment.)
For the third task, the Printed Form task, the answer sheet that was provided contained, in printed form, just the segments containing anomalies, along with the information that each segment printed on the form contained one and only one "mistake". This task allowed the elimination of some specious reactions in the first two tasks. That is, if a participant turned out to be unable to find and correct the the anomaly in the third task after having reacted to the segment containing it in the first or second task, then the reaction in the earlier task or tasks was treated as invalid. The Printed Form task was intended to provide maximum scope for participants to exploit L2 metalinguistic, analytical "knowledge of the rules". The answer sheet used for the third task also included questions at the bottom regarding personal background.