The present study looked at the internal reliability of the PEAK Relational Training Assessment, using a split-half method of measurement. The reliability of the assessment questions within each of the four factors, within the PEAK Relational Training Assessment was estimated through this process. Eighteen participants, between the ages of 26 months and ten years old were included in the study. All participants had been diagnosed with either a language based or developmental disability, including autism, seizure disorder, Down syndrome and related language disorders. The PEAK Relational Training Assessment (PEAK-D) was administered by a direct-care provider for each of the 18 participants and during standard instructional periods. Results indicate that for each of the 18 participants, there was a strong correlation between scores when one half of the items in each factor were compared to the remaining half. Results did show internal reliability for the PEAK-D when using split-half methodology.

Concerns regarding Mean Length of Utterance (MLU) has led to adaptations of this method of analysis. A recent study by Johnston (2001) introduced an alternative to MLU called MLU2. The current study investigates the split-half reliability of MLU and MLU2 as well as another alternative, Median Length of Utterance (Med-LU). Split-half reliability was found for these methods when segmented into Phonological and Communication Units. Split-half reliability of MLU2 was generally higher than that of MLU, and both were higher than Med-LU. The study suggests that MLU2 may also be a valuable tool for clinicians in analyzing child language.

In this paper, the approach of utilizing the core tensor from the Tucker decomposition, in place of theuncompressed tensor, for nding a valid tensor rank for the PARAFAC decomposition is considered.Validity of the proposed method is investigated in terms of error and time consumption. As thesolutions of the PARAFAC decomposition are unique, stability of the solutions through split-halfanalysis is investigated. Simulated and real data are considered. Although, no general validity ofthe method could be observed, the results for some datasets look promising with 10% compressionin all modes. It is also shown that increased compression does not necessarily imply less timeconsumption.

The approach-avoidance task (AAT) is an implicit task that measures people’s behavioral tendencies to approach or avoid stimuli in the environment. In recent years, it has been used successfully to help explain a variety of health problems (e.g., addictions and phobias). Unfortunately, more recent AAT studies have failed to replicate earlier promising findings. One explanation for these replication failures could be that the AAT does not reliably measure approach-avoidance tendencies. Here, we first review existing literature on the reliability of various versions of the AAT. Next, we examine the AAT’s reliability in a large and diverse sample (N = 1077; 248 of whom completed all sessions). Using a smartphone-based, mobile AAT, we measured participants’ approach-avoidance tendencies eight times over a period of seven months (one measurement per month) in two distinct stimulus sets (happy/sad expressions and disgusting/neutral stimuli). The mobile AAT’s split-half reliability was adequate for face stimuli (r = .85), but low for disgust stimuli (r = .72). Its test–retest reliability based on a single measurement was poor for either stimulus set (all ICC1s < .3). Its test–retest reliability based on the average of all eight measurements was moderately good for face stimuli (ICCk = .73), but low for disgust stimuli (ICCk = .5). Results suggest that single-measurement AATs could be influenced by unexplained temporal fluctuations of approach-avoidance tendencies. These fluctuations could be examined in future studies. Until then, this work suggests that future research using the AAT should rely on multiple rather than single measurements.