Triad-Comparison Task with Nonpolycomponential and Polycomponential Signs. 134

In Experiment 4 – a triad-comparison task with nonpolycomponential and polycomponential signs- addresses the same research question as Experiment 3. It explores whether the subjective judgment of meaning similarities between polycomponential actions and attributes corresponds with the results found in the verification task. The main question addressed is whether deaf signers judge polycomponential action- and attribute-related signs as stronger related to a noun referent than nonpolycomponential actions and attributes.

Method

Participants

The participants were the same 22 deaf (group 1), who performed in experiment 1, 2 and 3. The test-latency between Experiment 1 + 3 and Experiment 2 + 4 was two weeks.

Stimulus materials

The experimental materials overlapped those of Experiment 1. By removing the unrelated items (distractors), the test list contained 20 basic level target items (stimulus 1) combined randomly with 2 (stimulus 2+3) of 4 related items (nonpolycomponential action, nonpolycomponential attribute, polycomponential action and polycomponential attribute, i.e. 6 combinations within one set. Every possible combination of the items within a taxonomy was realized. Thus, there were 120 combinations of a target item with two related items. The presentation position was completely counterbalanced.

Table III.31: Example stimulus set (1 out of 20) of a basic level item in combination with two related items

Taxonomy Stimulus 1 Stimulus 2 Stimulus 3

Related Noun book (Buch)

Nonpoly-Action + Nonpoly-Attribute buy (kaufen) new (neu)

Nonpoly-Action + Poly-Attribute buy (kaufen) thick (dick)

Poly-Action + Nonpoly-Action read (lesen) buy (kaufen)

Poly-Action + Poly-Attribute read (lesen) thick (dick)

Nonpoly-Attribute + Poly-Action new (neu) read (lesen)

Poly-Attribute + Nonpoly-Attribute thick (dick) new (neu)

(see Appendix C for the whole list of items)

Procedure

The procedure in Experiment 4 was exactly the same as in Experiment 2. Deaf participants were asked to decide which of two presented items had a stronger semantic relation to another presented target item. The number of choices in favor to a specific semantic relation was measured for each group of participants. It was expected that the responses in the triad-comparison would correspond to the responses in the verification task for each group of participants, i.e. the lower the mean response time for polycomponential/ nonpolycomponential actions and attributes, respectively, the higher the relative number of choices for these groups of items.

Results

Two separate 2 x 2 ANOVAs were conducted on participants and items. The pre-analysis of the data is similar to that of the previous experiment. Each participant's mean error rates (missings) and number of choices in % for correct trials were determined for each condition. The same 12 items (3 in each group of items) which were removed from the item list in the previous experiment, were removed in the Triad-Task too. Error analysis were not significant and are not reported. Outliers were removed from the data. This procedure eliminated less than 4% of the data. The mean number of choices in % for correct trials are given in Table III.33:

Table III.32: Mean number of choices in % for correct trials and Standard Deviations (in brackets) for deaf participants in Experiment 3

A 2 x2 ANOVA was performed on the mean latency per relation category of deaf participants: 2 'class of semantic relations' x 2 'componential type’ ANOVA (see Table III.34).

Table III.33. 2 x 2 ANOVA/minF': Main effects and interaction

Effects Deaf Participants

Componential Type minF'(1,34)= 19.15, p = .0002

Class of Semantic Relation minF'(1,31)= 0.984, p = .3287 n.s.

Componential Type x Class of Semantic Relation minF'(1,30)= 0.028, p = .866 n.s.

Comparable with the results in the verification task with nonpoly- and polycomponential signs, the main effect of the variable 'Componential Type' was significant. The t-test for paired samples revealed a difference between mean number of choices of nonpolycomponential actions and polycomponential actions (t(21)=-11,592; p= .0001). The mean number of choices for polycomponential actions were 20 % higher than for nonpolycomponential actions. The difference between mean number of choices of nonpolycomponential attributes and polycomponential attributes (t(21)=-19,684; p= .0001) was also significant. Mean number of choices for polycomponential attributes were 24 % higher than for nonpolycomponential attributes. In general deaf participants chose more to polycomponential signs compared to nonpolycomponential signs.

Figure III.23: Mean Number of Choices % for each class of semantic relation (action and attribute) and componential type of item (nonpoly- and polycomponential)

The main effect of the variable 'class of semantic relations' and the interaction of the two factors were not significant. Comparable with the results in the verification task the mean number of choices in % for action-items was similar to the number of choices for attribute-items. However, the more detailed analysis revealed a difference between mean number of choices of nonpolycomponential actions and attributes (t(21)=7,187; p= .0001). Mean number of choices for nonpolycomponential actions were 8 % higher than for nonpolycomponential attributes. The difference between mean number of choices of polycomponential actions and attributes was also significant (t-test for paired samples, t(21)=2,740; p= .0012) (Bonferroni correction: p< 0.025).

Triad-Comparison Task with

Nonpolycomponential and Polycomponential Signs

0 10 20 30 40 50 60

Nonpolycomponential Polycomponential

Componential Type

Number of Choices in %

Action Attribute

Choices for nonpolycomponential actions were 4 % higher than for nonpolycomponential attributes. In detail deaf participants chose more action-signs compared to attribute-signs. This result confirms the outcome in Experiment 1.

With this outcome some additional evidence was found for the hypothesis that polycomponential signs have a stronger semantic relation to a noun-referent than nonpolycomponential signs in the semantic lexicon of deaf participants. The semantic relation between a polycomponential sign and a noun referent seems to be stronger compared to the semantic relation between a nonpolycomponential sign and a noun referent. The latter is confirmed not only by faster responses to, but also by a higher number of choices for polycomponential signs.

2.5. Measurement of Nonpolycomponential and Polycomponential Signs related to a Noun Referent in a Recognition Memory Task

Studies of memory errors revealed that when participants study lists of words, they think of semantic associates to those words during study and as a consequence, they make false recognitions on the basis of this prior activation ('False Recognition Effect' see Deese 1959;

Underwood 1965, Anisfeld & Knapp 1968; Hall & Kozloff 1973; Roedinger & McDermott 1995;

Robinson & Roediger 1997; Seamon, Luo & Gallo 1998; Seamon, Luo, Schlegel, Greene &

Goldenberg 2000).

The goal of Experiment 5 was to study false memory for associated signs of German Sign Language and to determine if deaf signers would made false recognitions of signs that were semantic associates of previously presented signs. Since false recognition is influenced by the strength of semantic relatedness between items it was hypothesized that the highest false memory would occur for polycomponential associates, because they contain a meaningful component that expresses an object characteristic of the previously studied sign-referent and therefore have a stronger semantic relation to a noun referent than nonpolycomponential associates. The main question addressed was, whether signers in a recognition memory task falsely recognize more polycomponential signs than nonpolycomponential signs.

Method

Participants

The participants were 20 deaf signers out of the group, who performed in experiment 1, 2 and 3.

Two signers were not able to attend the test at the obligatory testing-date. The test-latency between Experiment 1 + 3, Experiment 2 + 4 and Experiment 5 was about two weeks.

Stimulus Materials

The experimental materials consisted of four study-lists, each containing 20 target items and the accompanying four recognition-lists, each containing the 20 previously studied target items and in addition 40 nonstudied items. The nonstudied material consisted of 20 signs strongly associated with the studied signs and 20 non-related distractors. Again, the associated signs were divided into two groups of 10 items, i.e. polycomponential and nonpolycomponential associates (see Table 2.33). The studied items as well as the unstudied items were a balanced mixture of signs for objects, actions and attributes. The presentation position of target signs, polycomponential signs, nonpolycomponential signs and distractors in the recognition list was completely counterbalanced.

Table III.34: Example stimulus set (2 out of 4x20) of target level item in combination with 1.

polycomponential sign and distractor and 2. a nonpolycomponential sign and distractor

Target Polycomponential

Associate

Nonolycomponential Associate

Distracter

Glas (glass) spülen (wash up) Feder (feather)

bügeln (iron) Hemd (shirt) geheim (secret)

(see Appendix E for the whole list of items)

Procedure

In Experiment 5 a group of deaf participants was presented in four different sessions with a list of 20 signs to study. Subsequently, after the presentation of the list, they were asked whether the signs in a second list (60 items) occurred in the first list. The second list contained the 20 target signs, i.e. the signs presented in the first list, 10 polycomponential and 10 nonpolycomponential associates of the target signs and 20 distracters. The number of false positives and mean reaction times for false positives was measured. It was expected that the probability of false recognition of a nonstudied sign would increase with its associative strength to a studied sign. Thus the number of false positives for polycomponential signs should be high compared to nonpolycomponential signs and distractors.

The participants were tested in four different sessions at regular intervals of approximately 30 minutes. The standardized test-instructions were computer-based and informed the participants that they were taking part in a memory recognition experiment and that they would see 20 different signs, to which they should pay close attention because they were going to be asked to recognize them later. The study-session and recognition-session were demonstrated and the participants were given a practice trial. After the instruction and practice trials participants were informed that they had to press the space bar to begin presentation on the first study list.

The signs of the first list were displayed in a continuous sequence in the center of the computer screen and remained corresponding to their video-length. Each sign was followed by a blank screen for 2 sec before the next item appeared. Following the presentation of the final word in the list, the participants were asked to solve a set of four simple two-digit addition problems.

After approximately 30 seconds a yes/no recognition test was administered to each participant.

The recognition test items were presented individually on the computer screen and remained visible corresponding to their video length. Each video was followed by a blank screen for 2 seconds. The participants were instructed to press the right button of a 'Game Port Checker' if he or she thought the presented sign was identical with a sign studied before and the left button if he/she thought it was a new sign. The participants were told to do so as accurately and as quickly as possible at any time from the start of the first test item, but fast responses did not alter the timing. After each session the participants had a rest-period for 30 minutes.

Results

The pre-analysis of the data is similar to the ones of the previous experiments. Each participant's mean error rates (false positives and missings) and response times for false positives were determined for all sessions. One participant, 2 polycomponential, 3 nonpolycomponential and 10 distracter signs were excluded from the data, because of high missing rates.

Two separate ANOVAs were conducted on participants and items. Based on these two analysis, min F' was calculated. The one-way ANOVA was conducted on the crossed within-subject factors 'type of associative' with three levels (polycomponential, nonpolycomponential and non-related distractor). Dependent variables were Number of False Positives in % (fp in %) and Response Time in Milliseconds (RT in ms). Two separate one-way ANOVAs were conducted on the two factors. In addition multiple comparisons were performed in order to establish which pairs of levels are actually significant (Bonferroni procedure). An overview of the data is given in Table III.36.

Table III.35: False Positives in % and Response Time in ms for deaf signers in the recognition memory task Polycomponential Nonpolycomponential Non-Related Distracter

False Positives 23% (15) 11% (9) 9% (10)

Response Time 1669 ms (192) 1537 ms (183) 1498 ms (187)

(see Appendix E for detailed information)

A one-way repeated measures ANOVA of mean 'False positives in %' revealed a significant effect.

The influence of 'Type of Associative' on 'Response Time' was also significant (see Table III.37).

Table III.36: One-way repeated measures ANOVA/minF'

Effects False Positives in % Response Time

Type of Associative minF'(2,175)= 10.06, p = .0001 minF'(2,174)= 6.454, p = .002

Figure III.24. Mean False Positives in % of Deaf Participants in the Memory Recognition Task with Polycomponential and Nonpolycomponential Signs

Memory Recognition Task (Mean False Positives)

0 10 20 30 40 50

Polycomponential Nonpolycomponential Distracter

Type of Associate

% of False Positives

Figure III.25. Mean Response Time in ms for false positives of Deaf Participants in the Memory Recognition Task with Polycomponential and Nonpolycomponential Signs

An integrated chart comparing the mean score of 'False Positives' and 'Response Time' in each level is shown below (Figure III.26). On the first glance the distribution of errors and response times seem to be different for all levels of the factor 'Type of Associative', i.e. the largest number of errors and highest response times for polycomponential signs, followed by nonpolycomponential signs and finally distractors.

Pairwise comparisons of the differences in the means while adjusting the significance level using the Bonferroni procedure were performed in order to reveal which pairs of levels are actually significantly different. The number of errors as well as the response times for level one,

‘polycomponential’, was significantly different to all other levels (p.=0001). Non-significant differences were found between level two and three, ‘nonpolycomponential’ and ‘non-related distractor’, although the analyses of 'Response Time' revealed a trend for a significant difference between level 2 and 3 (p=.057).

Memory Recognition Task (Mean Response Time in ms)

1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000

Polycomponential Nonpolycomponential Non-Related (Distracter) Type of Associate

Response Time in ms

Figure III.26: Integrated Chart of Mean False Positives in % and Mean Response Time in % of Deaf Participants in the Memory Recognition Task with Polycomponential and Nonpolycomponential

Signs

In conclusion the experimental hypothesis that 'Type of Associative' has a significant effect on overall accuracy and response time and that in particular the associative type of level one 'Polycomponential' significantly decreases accuracy and increases response time over all other types of information is accepted.